Queue processing method and device, computer readable storage medium and equipment

Abstract

The embodiment of the invention discloses a queue processing method and device, a computer readable storage medium and equipment, and the method comprises the steps: obtaining promotion bit data of a promotion bit, carrying out the binning processing according to the promotion bit data, and obtaining the value data of the promotion bit after the binning processing; obtaining target object data of the target object, and performing binning processing according to the target object data to obtain object quality data after binning processing; combining the promotion bit value data and the object quality data to obtain target combination data; performing binning processing according to the target combined data to obtain target value data after binning processing; and based on the target binning interval in which the target value data is located, obtaining an adjustment weight of the target object under the promotion bit, and determining a corresponding candidate queue length based on the adjustment weight. Therefore, the data combinations of different dimensions are subjected to candidate queue length processing, the characteristics of the data of different dimensions can be integrated, the accuracy of computing resource allocation is improved, and the utilization rate of computing resources is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a queue processing method, apparatus, computer-readable storage medium and device. Background Art

[0002] At present, with the rapid development of terminal technology, applications have become an important channel for promoters to place promotional information with their advantages such as wide audience and high communication efficiency. Promoters can pay for promotion and place their own promotional information on the promotional positions of various applications, thereby promoting their own products and forming a better brand effect.

[0003] In the related art, for a request to place promotional information on a promotional position, the background needs to screen multiple promotional information in the promotional information library at multiple stages. A certain number of promotional information (i.e., candidate queues) will be screened out at each stage, and finally the most suitable promotional information will be selected for placement.

[0004] The length of the candidate queue allocated at each stage determines the size of the computing resources needed by the background, and will also affect the accuracy of the screened promotion information, and ultimately affect the delivery effect. However, in order to control the corresponding computing cost of computing resources, the length of the candidate queue cannot be allocated too long. Generally, the computing resources of each promotion position are reasonably set based on the value of the promotion position, and then the length of the candidate queue of each promotion position in each stage is allocated. Since this setting method refers to a single dimension and does not take into account the differences between different target objects under the promotion position, the computing resource allocation is unreasonable, resulting in a waste of computing resources. Summary of the invention

[0005] The embodiments of the present application provide a queue processing method, apparatus, computer-readable storage medium and device, which can improve the accuracy of computing resource allocation and thereby improve the utilization rate of computing resources.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] A queue processing method, comprising:

[0008] Acquire promotion position data of the promotion position, and perform binning processing according to the promotion position data to obtain promotion position value data after binning processing;

[0009] Acquire target object data of the target object, and perform binning processing according to the target object data to obtain object quality data after binning processing;

[0010] Combining the promotion position value data and the object quality data to obtain target combination data;

[0011] Perform binning processing according to the target combination data to obtain target value data after binning processing;

[0012] Based on the target binning interval in which the target value data is located, the adjustment weight of the target object under the promotion position is obtained, and the corresponding candidate queue length is determined based on the adjustment weight.

[0013] A queue processing device, comprising:

[0014] A first binning processing unit is used to obtain promotion position data of a promotion position, and perform binning processing according to the promotion position data to obtain promotion position value data after binning processing;

[0015] A second binning processing unit is used to obtain target object data of the target object, and perform binning processing according to the target object data to obtain object quality data after binning processing;

[0016] A combining unit, used for combining the promotional position value data and the object quality data to obtain target combination data;

[0017] A third binning processing unit, configured to perform binning processing according to the target combination data to obtain target value data after binning processing;

[0018] A determination unit is used to obtain an adjustment weight of the target object under the promotion position based on the target binning interval in which the target value data is located, and determine a corresponding candidate queue length based on the adjustment weight.

[0019] In some embodiments, the first binning processing unit includes:

[0020] The acquisition subunit is used to obtain the consumption data of the promotional position and the number of promotional position requests;

[0021] A calculation subunit, used for calculating the cost per thousand requests data corresponding to the promotion position according to the consumption data and the promotion position request amount;

[0022] A binning processing subunit, used for using the thousand request cost data as promotion position data, and performing equal frequency binning processing according to the promotion position data;

[0023] A determination subunit is used to determine the promotion position value data of the promotion position according to the promotion position bin interval in which the promotion position data is located.

[0024] In some embodiments, the binning processing subunit is used to:

[0025] Using the thousand request cost data as promotion position data;

[0026] Acquire first exposure data associated with the promotion position data, and perform statistics on the first exposure data to obtain first total exposure data;

[0027] Determine a first exposure bin threshold point according to the first total exposure data and a first preset bin number;

[0028] Based on the first exposure binning threshold point, the promotion position data is subjected to equal-frequency binning processing.

[0029] In some embodiments, the binning processing subunit is further used for:

[0030] Using the thousand request cost data as promotion position data;

[0031] Acquire first exposure data associated with the promotion position data, and perform statistics on the first exposure data to obtain first total exposure data;

[0032] Determine a first exposure bin threshold point according to the first total exposure data and a first preset bin number;

[0033] Sorting the promotion position data in ascending order;

[0034] Accumulate the first exposure data associated with the sorted promotion position data one by one according to the sorting, to obtain the first accumulated exposure data corresponding to each sorted promotion position data;

[0035] The promotion position data corresponding to the first accumulated exposure data that first reaches the first exposure bin threshold point is used as the promotion position bin interval threshold, and the promotion position bin interval is determined according to the promotion position bin interval threshold;

[0036] The promotion position data is binned based on the promotion position binning intervals.

[0037] In some embodiments, the second binning processing unit includes:

[0038] An acquisition subunit, used to acquire cost per thousand impressions data of a target object;

[0039] A binning processing subunit, used for taking the thousand exposure cost data as target object data, and performing equal-frequency binning processing according to the target object data;

[0040] A determination subunit is used to determine the object quality data of the target object according to the target object binning interval in which the target object data is located.

[0041] In some embodiments, the binning processing subunit is used to:

[0042] Using the cost per thousand impressions data as target object data;

[0043] Acquire second exposure data associated with the target object data, and perform statistics on the second exposure data to obtain second total exposure data;

[0044] Determine a second exposure bin threshold point according to the second total exposure data and a second preset bin number;

[0045] Based on the second exposure binning threshold point, equal-frequency binning is performed on the target object data.

[0046] In some embodiments, the binning processing subunit is further used for:

[0047] Using the cost per thousand impressions data as target object data;

[0048] Acquire second exposure data associated with the target object data, and perform statistics on the second exposure data to obtain second total exposure data;

[0049] Determine a second exposure bin threshold point according to the second total exposure data and a second preset bin number;

[0050] Sorting the target object data in ascending order;

[0051] Accumulating the second exposure data associated with the sorted target object data one by one according to the sorting, to obtain second accumulated exposure data corresponding to each sorted target object data;

[0052] Using the target object data corresponding to the second accumulated exposure data that first reaches the second exposure binning threshold point as the target object binning interval threshold, and determining the target object binning interval according to the target object binning interval threshold;

[0053] The target object data is binned based on the target object binning interval.

[0054] In some embodiments, the combination unit is used to:

[0055] Combining the promotion position value data with the first weight to obtain target promotion position value data;

[0056] Combining the object quality data with the second weight to obtain target object quality data;

[0057] The target promotion position value data and the target object quality data are summed to obtain target combination data.

[0058] In some embodiments, the device further comprises:

[0059] A load rate acquisition unit, used to acquire a target background load rate corresponding to the promotion position in a preset time period;

[0060] The combined unit is also used for:

[0061] Summing the promotion position value data and the object quality data to obtain intermediate combination data;

[0062] The ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

[0063] In some embodiments, the load rate acquisition unit is further used to:

[0064] Obtaining a first background load rate of the promotion position in a preset time period and a second background load rate in a historical preset time period;

[0065] Combining the first background load rate with the third weight to obtain a first target background load rate;

[0066] Combining the second background load rate with the fourth weight to obtain a second target background load rate;

[0067] The first target background load rate and the second target background load rate are summed to obtain the target background load rate of the promotion position.

[0068] In some embodiments, the combination unit is further used for:

[0069] Combining the promotion position value data with the fifth weight to obtain intermediate promotion position value data;

[0070] combining the object quality data and the sixth weight to obtain intermediate object quality data;

[0071] Summing the intermediate promotion position value data and the intermediate object quality data to obtain intermediate combination data;

[0072] The ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

[0073] In some embodiments, the third binning processing unit is used to:

[0074] Perform binning processing according to the target combination data;

[0075] Based on the target binning interval in which the target combination data is located, target value data of the target combination data is determined.

[0076] In some embodiments, the determining unit is configured to:

[0077] Based on the target binning interval in which the target value data is located, obtaining the adjustment weight of the target object under the promotion position;

[0078] Obtaining the initial queue length under the promotion position;

[0079] The product of the initial queue length and the adjustment weight is calculated to determine the candidate queue length of the target object under the promotion position.

[0080] The embodiment of the present application obtains the promotion position data of the promotion position, and performs binning processing according to the promotion position data to obtain the promotion position value data after binning processing; obtains the target object data of the target object, and performs binning processing according to the target object data to obtain the object quality data after binning processing; combines the promotion position value data and the object quality data to obtain the target combination data; performs binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, obtains the adjustment weight of the target object under the promotion position, and determines the corresponding candidate queue length based on the adjustment weight. In this way, the promotion position value data of the promotion position and the object quality data of the target object are combined to process the candidate queue length, so as to realize the allocation of computing resources taking into account the data characteristics of the promotion position and the data characteristics of the target object at the same time. Compared with the scheme in the related art that only sets computing resources according to the promotion position value of the promotion position, the embodiment of the present application greatly improves the accuracy of computing resource allocation, thereby improving the utilization rate of computing resources.

[0081] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0083] Figure 1 It is a structural diagram of the delivery system provided in an embodiment of the present application;

[0084] Figure 2 It is a scenario diagram of a queue processing system provided in an embodiment of the present application;

[0085] Figure 3 A schematic diagram of a process flow of a queue processing method provided in an embodiment of the present application;

[0086] Figure 4 A schematic diagram of another scenario of the queue processing method provided in an embodiment of the present application;

[0087] Figure 5 yes Figure 3 A specific flow chart of step 204;

[0088] Figure 6 yes Figure 3 Another specific flow chart of step 204;

[0089] Figure 7 Another schematic diagram of a process flow of a queue processing method provided in an embodiment of the present application;

[0090] Figure 8 A schematic diagram of another scenario of the queue processing method provided in an embodiment of the present application;

[0091] Fig. 9 It is a structural diagram of a queue processing device provided in an embodiment of the present application;

[0092] Fig.10 It is a schematic diagram of the structure of the server provided in the embodiment of the present application. DETAILED DESCRIPTION

[0093] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0094] It can be understood that in the specific implementation of the present application, related data such as promotion location data, target object data and consumption data are involved. When the above embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0095] It should be noted that in some processes described in the specification, claims and the above-mentioned drawings, multiple steps appearing in a specific order are included, but it should be clearly understood that these steps may not be executed in the order in which they appear in this document or may be executed in parallel. The step numbers are only used to distinguish different steps, and the numbers themselves do not represent any execution order. In addition, descriptions such as "first", "second" or "target" in this document are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0096] Before further describing the embodiments of the present disclosure in detail, the nouns and terms involved in the embodiments of the present disclosure are described. The nouns and terms involved in the embodiments of the present disclosure are subject to the following interpretations:

[0097] Promotional information, which can also be understood as advertising information, is the content conveyed by the advertiser to the advertising audience, that is, the content conveyed by the advertising source to the advertising destination through the advertising media. The content is in various forms (including text, images, audio, video, etc.).

[0098] Promotional position, which can be understood as an advertising position, is a window, control or component in an application used to display promotional information. When the target object operates the application and triggers the display of the promotional position, the promotional information after background screening will be displayed on the promotional position.

[0099] Consumption data refers to the revenue of the promotion position, such as the average daily revenue of the promotion position. The source of these revenues is the consumption of the promotion information released by the promoter.

[0100] Click-through rate (CTR) refers to the ratio of the number of times the promotional information on the promotional position is clicked to the number of times it is displayed.

[0101] Conversion rate refers to the ratio of the number of target objects who purchase, register or demand information due to the influence of promotional information to the total number of people who click on the promotional information.

[0102] Cost Per Mille (CPM) is calculated based on the impressions, which is the cost of showing the promotion information to one thousand people, so it is also called the cost per thousand impressions. For example, if the cost of a promotion position is 8, then the CPM is 8.

[0103] Binning, a part of data preprocessing, is also called bucketing or discretization. It can group data intervals into intervals ("boxes" or "buckets") to better understand the distribution of data.

[0104] In equal-frequency binning, the number of data contained in each interval ("box" or "bucket") after grouping is equal, that is, the amount of data included in each interval is the same.

[0105] In the equidistant binning process, the distance between the values ​​of each two intervals is the same, and the data is classified according to the equidistant intervals.

[0106] At present, when the target object operates the application and triggers the display of the promotion position, a request for placing promotion information on the displayed promotion position is generated, that is, the adapted promotion information is displayed on the displayed promotion position, so that the target object clicks to understand the promotion information, and even consumes the product behind the promotion information, achieving a better promotion effect. For example, when the target object uses a video application to open a video, before the video is played, it will trigger the display of the promotion position on the video playback window, and the video application will filter out the promotion information adapted to the target object in the background for display.

[0107] It is understandable that there are often multiple promoters who will promote and pay for one promotional position at the same time. Since each promoter has different brands, purposes and products, the corresponding promotional information delivered is also different. Therefore, each promotional position will store multiple promotional information in the background promotion information library.

[0108] At the same time, different target objects often have different preferences for promotional information due to their different era backgrounds, educational backgrounds, and cultural backgrounds. In actual scenarios, target objects often prefer promotional information of the type they like, and are more likely to click on and consume promotional information of the type they like. Therefore, in order to achieve better delivery effects, the promotion position needs to screen the promotional information for each target object, and select the most appropriate promotional information from the background promotion information library for delivery to achieve better delivery effects.

[0109] Please also read Figure 1 As shown, Figure 1 It is a structural diagram of the delivery system provided in the embodiment of the present application, that is, each promotion position selects the most suitable promotion information from the background promotion information library for delivery, and will go through three stages of screening: recall A, rough sorting B, and fine sorting C, to obtain the most suitable promotion information to realize the stage of promotion information display D, thereby realizing the delivery of promotion information.

[0110] Recall Phase A: The goal of Recall Phase A is to recall some promotional information (for example, thousands of promotional information) from a promotional information database of tens of millions or even hundreds of millions. Recall Phase A generally consists of multiple paths, each of which has a different focus (optimization goal).

[0111] In the rough sorting stage B, the recall results of the recall stage A are roughly sorted to reduce the data size of the promotion information (for example, reducing thousands of promotion information to hundreds of promotion information).

[0112] In the fine sorting stage C, after the screening in the rough sorting stage B, since there are very few candidate promotion information at this time, the fine sorting stage C can accurately screen the rough sorting results of the rough sorting stage B to obtain the most accurate promotion information and realize the promotion information display stage D.

[0113] It can be seen that both the recall stage A and the rough sorting stage B will screen out a certain amount of promotion information, which will serve as the data set to be screened in the next stage. For example, the certain amount of promotion information recalled in recall A will serve as the data set screened in the rough sorting stage B, and the certain amount of promotion information screened in the rough sorting stage B will serve as the data set screened in the fine sorting stage C.

[0114] Therefore, a certain number of promotion information screened out in the recall stage A and the rough sorting stage B is the candidate queue. For example, a certain number of promotion information screened out in the recall stage A is the candidate queue for the rough sorting stage B, and a certain number of promotion information roughly sorted out in the rough sorting stage B is the candidate queue for the fine sorting stage C.

[0115] It should be noted that in order to avoid the situation where the ad slot has been displayed but the promotional information has not been delivered for a long time, resulting in a blank ad slot and affecting the user experience, it is necessary to ensure the real-time delivery of promotional information on each promotion slot. That is, each promotion slot needs to display the screened promotional information in real time within a set time from the time it is displayed. For example, the promotion slot can be set to display the screened promotional information in real time within 4 milliseconds from the time it is displayed. From this, it can be seen that the total computing time for promotional information screening is set, and accordingly, the computing time for each stage is also set. For example, the computing time for recall stage A can be 1 millisecond, the computing time for rough sorting stage B can be 1 millisecond, and the computing time for fine sorting stage C can be 2 milliseconds.

[0116] Therefore, in the scenario where the computing time of each stage is set, the length of the candidate queue directly affects the size of the computing resources that need to be allocated in the background. For example, the longer the length of the candidate queue, the more promotion information that needs to be screened. Correspondingly, if the length of the candidate queue is shorter, the less promotion information that needs to be screened. In the scenario where the computing time of each stage is set, the longer the length of the candidate queue, the more computing resources that need to be allocated, and the shorter the length of the candidate queue, the fewer computing resources that need to be allocated. The computing resources can be central processing unit (CPU) resources, memory resources, graphics processing unit (GPU) resources and other resources used to process computing tasks.

[0117] Accordingly, the length of the candidate queue not only determines the allocation of computing resources, but also affects the accuracy of the promotion information screened out, because the longer the candidate queue is, the more promotion information is involved in the calculation, and the more accurate the promotion information screened out at the corresponding stage is, while the shorter the candidate queue is, the less promotion information is involved in the calculation, and the less accurate the promotion information screened out at the corresponding stage is. And through the screening process at each stage, it will eventually be fed back to the final delivery effect.

[0118] Through the above analysis, it can be seen that increasing the length of the candidate queue can improve the accuracy of the promotion information screened out in the corresponding stage. However, it will also increase the computing resources and computing costs of the corresponding stage. Therefore, it is necessary to balance the relationship between the length of the candidate queue and the computing cost. In actual scenarios, due to the different promotion position values ​​of different promotion positions, promotion positions with high promotion position values ​​are often allocated more computing resources, and the length of their corresponding candidate queues will also be set longer. For promotion positions with low promotion position values, fewer computing resources will be allocated, and the length of their corresponding candidate queues will also be set shorter, so as to achieve flexible allocation of computing resources.

[0119] However, the method of allocating computing resources only by the value of promotion positions has a single reference dimension and does not take into account the differences between different target objects, that is, the different era backgrounds, educational backgrounds, and cultural backgrounds between target objects will lead to different demands and consumption capabilities for products. Therefore, there are also differences in the demand for promotion information between different target objects. Allocating computing resources only by the value of promotion positions will result in too few computing resources being allocated to target objects with high quality in promotion positions with low promotion position values, resulting in poor delivery effects, and will cause target objects with high quality to be dissatisfied with the delivery effects, and even lose target objects with high quality. Correspondingly, it will also result in too many computing resources being allocated to target objects with low quality in promotion positions with high promotion position values. Since the demand and consumption capabilities of target objects with low quality for products are limited, it will lead to a waste of computing resources. Therefore, the above-mentioned method of allocating computing resources based on the characteristics of single-dimensional data will make the allocation of computing resources unreasonable, resulting in a waste of computing resources.

[0120] The embodiments of the present application can allocate computing resources with reference to multi-dimensional data, thereby improving the accuracy of computing resource allocation and thus improving the utilization rate of computing resources. Please refer to the following embodiments for specific implementation methods.

[0121] See also Figure 2 , Figure 2The scenario diagram of the queue processing system provided for the embodiment of the present application includes: a terminal 110, and a server 120 (the queue processing system may also include other terminals 110 in addition to the terminal 110, and the specific number of the terminal 110 is not limited here), the terminal 110 and the server 120 may be connected through a communication network, and the communication network may include a wireless network and a wired network, wherein the wireless network includes a combination of one or more of a wireless wide area network, a wireless local area network, a wireless metropolitan area network, and a wireless personal network. The network includes network entities such as routers and gateways, which are not shown in the figure. The terminal 110 can exchange information with the server 120 through the communication network. For example, the terminal 110 can send a promotion information request to the server 120, and the server 120 can filter out the corresponding promotion information according to the promotion information request and return it to the terminal 110 for display.

[0122] The queue processing system may include a queue processing device, which may be specifically integrated in the server 120. The server 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Figure 1 As shown, the server 120 obtains promotion position data of the promotion position, and performs binning processing according to the promotion position data to obtain promotion position value data after binning processing; obtains target object data of the target object, and performs binning processing according to the target object data to obtain object quality data after binning processing; combines the promotion position value data and the object quality data to obtain target combination data; performs binning processing according to the target combination data to obtain target value data after binning processing; based on the target binning interval where the target value data is located, obtains the adjustment weight of the target object under the promotion position, and determines the corresponding candidate queue length based on the adjustment weight.

[0123] In the queue processing system, the terminal 110 can install various applications required by users, such as instant messaging applications, video applications, and browser applications. When the terminal 110 displays the promotion position, it will generate a promotion information request and send it to the server, so that the server 120 can filter out the corresponding promotion information according to the promotion information request and feed it back to the promotion position of the terminal 110 for display.

[0124] It should be noted that Figure 2The scenario diagram of the queue processing system shown is only an example. The queue processing system and scenario described in the embodiment of the present application are intended to more clearly illustrate the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. Ordinary technicians in this field can know that with the evolution of the queue processing system and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0125] In this embodiment, the description will be made from the perspective of a queue processing device, which can be specifically integrated in a computer device having a storage unit and a microprocessor installed therein and having computing power. The computer device can be a server or a terminal. In this embodiment, the computer device is described as a server.

[0126] See also Figure 3 , Figure 3 : is a flow chart of a queue processing method provided in an embodiment of the present application. The queue processing method includes:

[0127] In step 201, the promotion position data of the promotion position is obtained, and the promotion position data is binned to obtain the promotion position value data after binning.

[0128] Among them, the promotion position can be understood as an advertising position, which is a window, control or component in the application used to display promotional information (i.e., advertising information). When the target object operates the application and triggers the display of the promotion position, the promotion information filtered by the server will be displayed on the promotion position.

[0129] For example, see also Figure 4 As shown, when the target object is browsing the social circle interface in the instant messaging application and triggers the display of the promotion position 1101 on the social circle interface, the instant messaging application will display the promotion information filtered by the server 120 on the promotion position of the social circle interface; when the target object opens the video playback interface in the video application to play the video, before the video plays, it will trigger the display of the promotion position on the video playback window, and the video application will display the promotion information filtered by the server on the promotion position; when the target object is viewing the news interface in the browser application and triggers the display of the promotion position on the news interface, the browser application will display the promotion information filtered by the server on the promotion position of the news interface.

[0130] It can be seen that different promotion positions have differences in consumption, cost per thousand exposures, click-through rate and conversion rate. Therefore, the promotion position values ​​of different promotion positions are different. The higher the value of the promotion position, the better the promotion effect of the promotion position. The lower the value of the promotion position, the worse the promotion effect of the promotion position.

[0131] In the embodiment of the present application, the value of the promotion position can be measured by the promotion position value data. The higher the value of the promotion position value data, the higher the value of the promotion position, and the lower the value of the promotion position value data, the lower the value of the promotion position.

[0132] In order to obtain the promotion position value data, the embodiment of the present application can obtain the promotion position data of each promotion position in advance. The promotion position data can be data such as cost per thousand exposures or cost per thousand requests (CPR). The cost per thousand requests can be understood as the cost required for one thousand promotion position requests. The calculation formula is CPR (cost per thousand requests data) = consumption data / promotion position request amount * 1000. The promotion position request amount is the number of candidate queues requested for screening on the promotion position.

[0133] In this way, the cost per thousand impressions of each promotion position can be obtained as promotion position data, and the cost per thousand requests of each promotion position can also be obtained as promotion position data. Among them, the cost per thousand impressions data can reflect the value characteristics of each promotion position from the perspective of exposure cost. The higher the cost per thousand impressions data, the higher the value of the promotion position, and the lower the cost per thousand impressions data, the lower the value of the promotion position. The cost per thousand requests data can reflect the value characteristics of each promotion position from the perspective of request cost. The higher the cost per thousand requests data, the higher the value of the promotion position, and the lower the cost per thousand requests data, the lower the value of the promotion position.

[0134] Furthermore, since the promotion position data of different promotion positions are different, the differences between the promotion position data are relatively large. In order to obtain the promotion position value data of each promotion position, the promotion position data needs to be binned. In the embodiment of the present application, the promotion position data is binned by binning, that is, each promotion position data can be sorted in order from small to large, and the number of bins can be determined according to the number of bins required. The number of bins can be set according to actual needs, and there is no specific quantity limit here. The sorted promotion position data are binned in turn according to the number of bins, and each promotion position data is binned into a corresponding interval (i.e., "box") to achieve binning of all promotion position data. Each interval can correspond to a promotion position value data, thereby obtaining the promotion position value data after binning for each promotion position. The binning method can be equal-frequency binning or equal-distance binning.

[0135] For example, the number of bins is 3, which means there are 3 promotion position bin intervals. Bin 1 corresponds to the promotion position value data -1, bin 2 corresponds to the promotion position value data 0, and bin 3 corresponds to the promotion position value data 1. Based on this, all the promotion position data can be divided into 3 parts according to the sorting and assigned to the corresponding promotion position bin intervals to realize bin processing. According to the result of the bin processing, the promotion position value data of each promotion position data after binning is obtained, that is, the promotion position value data of each promotion position can be obtained, so as to realize the grading of the promotion position values ​​of different promotion positions.

[0136] It should be noted that using cost per thousand requests data as promotion position data can also introduce considerations of the computing resource dimension. For example, when the consumption data and exposure times of promotion position 1 and promotion position 2 are the same, assuming that the candidate queue of promotion position 1 is 100 (the number of promotion position requests is 100), and the candidate queue of promotion position 2 is 200 (the number of promotion position requests is 200). At this time, the cost per thousand exposures data of the two are the same. However, under the same cost per thousand exposures data, the computing resources required for promotion position 2 are twice that of promotion position 1, that is, the cost per thousand requests data of promotion position 1 is higher than the cost per thousand requests data of promotion position 2. Therefore, using cost per thousand requests data as promotion position data can also comprehensively consider the relationship between computing resources and costs.

[0137] In one implementation, taking the cost per thousand requests data as promotion position data as an example, step 201 includes:

[0138] (1) Obtaining consumption data and promotional position request volume of promotional positions;

[0139] (2) Calculate the cost per thousand requests corresponding to the promotional position based on the consumption data and the number of promotional position requests;

[0140] (3) using the thousand-times-requests cost data as promotion position data, and performing equal-frequency binning processing based on the promotion position data;

[0141] (4) Determine the promotion position value data of the promotion position according to the promotion position sub-box interval in which the promotion position data is located.

[0142] Among them, the consumption data and promotion position request volume of each promotion position within a cycle can be obtained. The promotion position request volume is the number of candidate queues requested for screening on the promotion position. For example, the offline statistics of the consumption data of promotion position 1 within a week are 10,000, the candidate queues requested for screening in the rough sorting stage are 20,000, and the candidate queues requested for screening in the fine sorting stage are 30,000. Then the promotion position request volume of promotion position 1 is 50,000.

[0143] Furthermore, according to the formula CPR (cost per thousand requests) = consumption data / promotional position request volume * 1000, the consumption data is divided by the promotional position request volume and then multiplied by 1000 to obtain the corresponding cost per thousand requests data for each promotional position. For example, through the above formula, the corresponding cost per thousand requests data for promotion position 1 within one week is 200.

[0144] In this way, the thousand-request cost data is used as the promotion position data, and the relationship between resources and costs can be comprehensively calculated. Since the thousand-request cost data of different promotion positions are different, for example, there are 9 promotion positions, namely promotion position 1, promotion position 2, promotion position 3, promotion position 4, promotion position 5, promotion position 6, promotion position 7, promotion position 8 and promotion position 9, the promotion position data (i.e., thousand-request cost data) of promotion position 1 is 200, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700, the promotion position data of promotion position 8 is 80, and the promotion position data of promotion position 9 is 900.

[0145] Therefore, the embodiment of the present application needs to perform equal-frequency binning processing on the promotion position data to achieve binning processing of promotion position data with different differences. In one implementation, the binning processing can be achieved by the following formula:

[0146] x = bin(CPR, bin_num = X)

[0147] The x is the promotion position value data, the bin() represents the binning function, and the X is the bin number. The bin number can be set according to actual needs. For example, 3 bins indicate that there are 3 promotion position bin intervals, and bin 1 corresponds to the promotion position value data -1, bin 2 corresponds to the promotion position value data 0, and bin 3 corresponds to the promotion position value data 1. Based on this, the 9 promotion positions are first sorted according to the promotion position data to obtain the sorted promotion position data, that is, the promotion position data of promotion position 8 is 80, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700, and the promotion position data of promotion position 9 is 900.

[0148] According to the sorting order, the promotion position data is divided into three parts according to the number of boxes, namely (the promotion position data of promotion position 8 is 80, the promotion position data of promotion position 2 is 150, and the promotion position data of promotion position 5 is 180), (the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230, and the promotion position data of promotion position 4 is 400) and (the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700, and the promotion position data of promotion position 9 is 900), and each part is assigned to the corresponding promotion position box interval, that is, (the promotion position data of promotion position 8 is 80, the promotion position data of promotion position 2 is 150, and the promotion position data of promotion position 5 is 180), (the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230, and the promotion position data of promotion position 4 is 400) The promotion position data of promotion position 2 is 150 and the promotion position data of promotion position 5 is 180) are classified into the first promotion position sub-box interval (i.e., sub-box 1), (the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230 and the promotion position data of promotion position 4 is 400) are classified into the second promotion position sub-box interval (i.e., sub-box 2), and (the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700 and the promotion position data of promotion position 9 is 900) are classified into the third promotion position sub-box interval (i.e., sub-box 3), thereby realizing equal-frequency binning processing according to the promotion position data.

[0149] Based on this, the promotion position value data corresponding to each promotion position data can be obtained according to the promotion position bin interval in which the promotion position data is located, that is, the promotion position value data of each promotion position can be obtained. For example, the promotion position value data of promotion position 8, promotion position 2 and promotion position 5 is -1, the promotion position value data of promotion position 1, promotion position 3 and promotion position 4 is 0, and the promotion position value data of promotion position 6, promotion position 7 and promotion position 9 is 1. That is to say, the promotion position value data of promotion position 8, promotion position 2 and promotion position 5 are the lowest level, the promotion position value data of promotion position 1, promotion position 3 and promotion position 4 are the middle level, and the promotion position value is higher than the lowest level, and the promotion position value data of promotion position 6, promotion position 7 and promotion position 9 are the highest level, and the promotion position value is higher than the middle level. In this way, the promotion position values ​​of different promotion positions are classified according to the promotion position data from high to low.

[0150] In some implementations, the equal-frequency binning process is performed according to the promotional position data, including:

[0151] (1.1) Obtaining first exposure data associated with the promotion position data, and performing statistics on the first exposure data to obtain first total exposure data;

[0152] (1.2) determining a first exposure bin threshold point according to the first total exposure data and a first preset bin number;

[0153] (1.3) Based on the first exposure binning threshold point, the promotion position data is subjected to equal-frequency binning processing.

[0154] Each promotion position also includes first exposure data (Page View, PV), which refers to the number of exposures of the promotion information exposed on the promotion position. Therefore, each promotion position includes both promotion position data and first exposure data, so that the promotion position data of each promotion position can be associated with the corresponding first exposure data. It should be noted that the first exposure data on different promotion positions are different. In order to achieve the balance of exposure data covered in each promotion position binning interval after equal-frequency binning processing, the embodiment of the present application can perform equal-frequency binning processing from the perspective of exposure data.

[0155] In this way, the first exposure data corresponding to the promotion position behind each promotion position data can be obtained, and the first exposure data of all promotion positions are counted to obtain the first total exposure data of all promotion positions. For example, there are M promotion positions, each of which contains a number of first exposure data. The first exposure data of the M promotion positions are counted to obtain the first total exposure data N of the M promotion positions. M and N can be positive integers.

[0156] Furthermore, the first preset bin number is the bin number, which can be set according to actual needs, for example, X bins. Thus, the first exposure bin threshold point of each promotion position bin interval can be determined according to the first total exposure data N and the first preset bin number, that is, ((N / X), 2*(N / X), ..., X*(N / X)). The first exposure bin threshold point is the bin critical point for equal-frequency binning of the promotion position data according to the angle of the exposure data.

[0157] In this way, the promotion position data are sorted in order from small to large, and the first accumulated exposure data of the promotion positions behind the promotion position data are accumulated in order from small to large. When the first accumulated exposure data reaches (N / X), the promotion position data of the corresponding promotion position a is the promotion position bin interval threshold of the first bin, that is, the promotion positions whose promotion position data is less than or equal to the promotion position data of promotion position a are all classified into the first bin.

[0158] Correspondingly, when the first accumulated exposure data reaches (2*(N / X)), the corresponding promotion position data of promotion position b is the promotion position bin interval threshold of the second bin, that is, the promotion position data that is greater than the promotion position data of promotion position a and less than or equal to the promotion position data of promotion position b are all classified into the second bin, and so on, each promotion position data is classified into the corresponding bin, and the promotion position data is processed with equal frequency bins according to the first exposure bin threshold point. This processing method performs equal frequency binning from the perspective of exposure data, which can make the data covered in each bin after the equal frequency binning processing more balanced, and improve the rationality of the subsequent promotion position value data setting.

[0159] In some implementations, performing equal-frequency binning processing on the promotion position data based on the first exposure binning threshold point includes:

[0160] (2.1) Sort the promotion position data in ascending order;

[0161] (2.2) accumulating the first exposure data associated with the sorted promotion position data one by one according to the sorting, to obtain the first accumulated exposure data corresponding to each sorted promotion position data;

[0162] (2.3) using the promotion position data corresponding to the first accumulated exposure data that first reaches the first exposure bin threshold point as the promotion position bin interval threshold, and determining the promotion position bin interval according to the promotion position bin interval threshold;

[0163] (2.4) The promotion position data is binned based on the promotion position binning interval.

[0164] Among them, the promotion position data can be first sorted in order from small to large. For example, there are 9 promotion positions, the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700, the promotion position data of promotion position 8 is 80, and the promotion position data of promotion position 9 is 900. The promotion position data are sorted in ascending order, that is, the promotion position data of promotion position 8 is 80, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700 and the promotion position data of promotion position 9 is 900 in this order.

[0165] Each promotion position includes both promotion position data and first exposure data, so that the promotion position data of each promotion position can be associated with the corresponding first exposure data. For example, the first exposure data of promotion position 1 is 40, the first exposure data of promotion position 2 is 30, the first exposure data of promotion position 3 is 20, the first exposure data of promotion position 4 is 40, the first exposure data of promotion position 5 is 50, the first exposure data of promotion position 6 is 40, the first exposure data of promotion position 7 is 40, the first exposure data of promotion position 8 is 10, and the first exposure data of promotion position 9 is 60. The first exposure data of all promotion positions are counted to obtain the first total exposure data of 330. In the embodiment of the present application, the number of the first preset bins is 3, so the first exposure bin threshold point is (110, 220 and 330). The first exposure bin threshold point is the critical point for equal-frequency binning of the promotion position data according to the angle of the exposure data.

[0166] Furthermore, the first exposure data associated with the sorted promotion position data can be accumulated one by one in the sorting order to obtain the first accumulated exposure data corresponding to each sorted promotion position data. For example, the first accumulated exposure data of the sorted promotion position 8 is 10, the first accumulated exposure data of the promotion position 2 is 40, the first accumulated exposure data of the promotion position 5 is 90, the first accumulated exposure data of the promotion position 1 is 130, the first accumulated exposure data of the promotion position 3 is 150, the first accumulated exposure data of the promotion position 4 is 190, the first accumulated exposure data of the promotion position 6 is 230, the first accumulated exposure data of the promotion position 7 is 270 and the first accumulated exposure data of the promotion position 9 is 330.

[0167] Thus, the promotion position data 200 of promotion position 1 that first reaches the first exposure bin threshold point 110 is used as the promotion position bin interval threshold of the first bin; the promotion position data 550 of promotion position 6 that first reaches the first exposure bin threshold point 220 is used as the promotion position bin interval threshold of the second bin; the promotion position data 900 of promotion position 9 that first reaches the first exposure bin threshold point 330 is used as the promotion position bin interval threshold of the third bin. Accordingly, the promotion position bin intervals ((0-200], (200-550], (550-900]) can be determined according to the promotion position bin interval thresholds 200, 550 and 900.

[0168] Based on the promotion position binning interval, promotion position 8, promotion position 2, promotion position 5 and promotion position 1 with promotion position data less than 200 are all classified into the first bin; promotion position 3, promotion position 4 and promotion position 6 with promotion position data greater than 200 and less than or equal to 550 are all classified into the second bin; promotion position 7 and promotion position 9 with promotion position data greater than 550 and less than or equal to 900 are all classified into the third bin, so as to realize binning processing of promotion position data, so that the data covered in each bin after equal-frequency binning processing is more balanced.

[0169] In step 202, target object data of the target object is acquired, and binning processing is performed according to the target object data to obtain object quality data after binning processing.

[0170] Among them, different target objects have different demands and consumption capabilities for products due to their different era backgrounds, educational backgrounds, and cultural backgrounds. Therefore, the object quality of different target objects is different. The higher the object quality, the greater the demand of the target object for the promoted product and the stronger the consumption capability. For target objects with high object quality, more computing resources can be allocated. On the contrary, the lower the object quality, the smaller the demand of the target object for the promoted product and the smaller the consumption. For target objects with low object quality, computing resources can be appropriately reduced.

[0171] In the embodiment of the present application, the object quality can be measured by the object quality data. The higher the value of the object quality data, the higher the quality of the object, and the lower the value of the object quality data, the lower the quality of the object.

[0172] In order to obtain object quality data, the embodiment of the present application can use target object data related to the promotion effect to measure the object quality of the target object, that is, the target object data of each target object can be obtained in advance. The target object data can be cost per thousand exposures data, number of clicks data, click-through rate data, number of conversions data or conversion rate data, etc.

[0173] The embodiment of the present application takes the target object data as the thousand exposure cost data as an example for explanation, and the thousand exposure cost data of each target object can be obtained as the target object data. In one embodiment, the average of the thousand exposure cost data of all promotional information exposed to the target object in a period (for example, 7 days, 15 days or 30 days) can be counted as the thousand exposure cost data of the target object, and the average of the promotional information of the target object in a period of a preset number (for example, 10, 20 or 30) can also be counted as the thousand exposure cost data of the target object to ensure the real-time nature of the thousand exposure cost data. The thousand exposure cost data can reflect the object quality characteristics of each target object from the perspective of exposure cost. The higher the thousand exposure cost data, the higher the object quality of the target object, and the lower the thousand exposure cost data, the lower the object quality of the target object.

[0174] Furthermore, since the target object data of different target objects are different, the differences between the target object data are relatively large. In order to obtain the object quality data of each target object, the target object data needs to be binned. In the embodiment of the present application, the target object data can be binned by binning, that is, each target object data can be sorted in order from small to large, and the number of bins can be determined according to the number of bins required. The number of bins can be set according to actual needs, and no specific quantity limit is made here. The sorted target object data is binned in turn according to the number of bins, and each target object data is binned into a corresponding interval to achieve binning of all target object data, and each interval can correspond to an object quality data, thereby obtaining the object quality data of each target object after binning. The binning method can be equal-frequency binning or equal-distance binning.

[0175] For example, if the number of bins is 3, it means that there are 3 target object binning intervals, bin 1 corresponds to object quality data -1, bin 2 corresponds to object quality data 0, and bin 3 corresponds to object quality data 1. Based on this, all target object data can be divided into 3 parts according to the sorting and assigned to the corresponding target object binning intervals to implement binning processing. According to the results of the binning processing, the object quality data of each target object data after binning is obtained, that is, the object quality data of each target object can be obtained, so as to implement grading of the object quality of different target objects.

[0176] In some implementations, step 202 includes:

[0177] (1) Obtain the cost per thousand impressions data of the target object;

[0178] (2) using the thousand-times-exposure cost data as target object data, and performing equal-frequency binning processing based on the target object data;

[0179] (3) Determine the object quality data of the target object according to the target object binning interval in which the target object data is located.

[0180] Among them, the average cost per thousand exposures data of the target object's most recent 50 promotional information exposures within a month can be obtained, and the average value can be used as the cost per thousand exposures data of the target object to ensure the real-time nature of the cost per thousand exposures data of each target object.

[0181] In this way, the cost per thousand exposures data is used as the target object data. Since the cost per thousand exposures data of different target objects are different, for example, there are 6 target objects, namely target object 1, target object 2, target object 3, target object 4, target object 5 and target object 6. The target object data of target object 1 (i.e., the cost per thousand exposures data) is 100, the target object data of target object 2 is 50, the target object data of target object 3 is 120, the target object data of target object 4 is 200, the target object data of target object 5 is 80, and the target object data of target object 6 is 300.

[0182] Therefore, the embodiment of the present application needs to perform equal-frequency binning processing on the target object data to achieve binning of target object data with different differences. In one implementation, the binning processing can be achieved by the following formula:

[0183] y = bin(CPM, bin_num = w)

[0184] The y is the object quality data, the bin() represents the binning function, and the w is the bin number. The bin number can be set according to actual needs. For example, 3 bins means there are 3 target object bin intervals, bin 1 corresponds to object quality data -1, bin 2 corresponds to object quality data 0, and bin 3 corresponds to object quality data 1. Based on this, the 6 target objects are first sorted according to the target object data to obtain the sorted target object data, that is, the target object data of target object 2 is 50, the target object data of target object 5 is 80, the target object data of target object 1 is 100, the target object data of target object 3 is 120, the target object data of target object 4 is 200, and the target object data of target object 6 is 300.

[0185] According to the sorting order, the target object data is divided into three parts according to the number of bins, namely (the target object data of target object 2 is 50, the target object data of target object 5 is 80), (the target object data of target object 1 is 100, the target object data of target object 3 is 120) and (the target object data of target object 4 is 200 and the target object data of target object 6 is 300), and each part is assigned to the corresponding target object bin interval, that is, (the target object data of target object 2 is 50, the target object data of target object 5 is 80) is assigned to the first target object bin interval (i.e. bin 1), (the target object data of target object 1 is 100, the target object data of target object 3 is 120) is assigned to the second target object bin interval (i.e. bin 2), and (the target object data of target object 4 is 200 and the target object data of target object 6 is 300) is assigned to the third target object bin interval (i.e. bin 3), so as to realize equal-frequency binning processing according to the target object data.

[0186] Based on this, the object quality data corresponding to each target object data can be obtained according to the target object binning interval in which the target object data is located, that is, the object quality data of each target object can be obtained. For example, the object quality data of target object 2 and target object 5 is -1, the object quality data of target object 1 and target object 3 is 0, and the object quality data of target object 4 and target object 6 is 1. That is, the object quality data of target object 2 and target object 5 is the lowest gear, the object quality data of target object 1 and target object 3 is the middle gear, and the object quality data is higher than the lowest gear, and the object quality data of target object 4 and target object 6 is the highest gear, and the object quality data is higher than the middle gear. In this way, the object quality data of different target objects are binned according to the order of target object data from high to low.

[0187] In some implementations, performing equal-frequency binning processing according to the target object data includes:

[0188] (1.1) obtaining second exposure data associated with the target object data, and performing statistics on the second exposure data to obtain second total exposure data;

[0189] (1.2) determining a second exposure bin threshold point according to the second total exposure data and a second preset bin number;

[0190] (1.3) Based on the second exposure binning threshold point, the target object data is subjected to equal-frequency binning processing.

[0191] Each target object also includes second exposure data, and the second exposure data refers to the exposure quantity of the promotional information exposed on the target object. Therefore, each target object includes target object data and second exposure data at the same time, so that the target object data of each target object can be associated with the corresponding second exposure data. It should be noted that the second exposure data on different target objects is different. In order to achieve the balance of exposure data covered in each target object binning interval after equal-frequency binning processing, the embodiment of the present application can perform equal-frequency binning processing from the perspective of exposure data.

[0192] In this way, the second exposure data corresponding to the target object behind each target object data can be obtained, and the second exposure data of all target objects are counted to obtain the second total exposure data of all target objects. For example, there are Y target objects, each of which contains a number of second exposure data. The second exposure data of the Y target objects are counted to obtain the second total exposure data P of the Y target objects. Y and P can be positive integers.

[0193] Furthermore, the second preset number of bins is the number of bins, which can be set according to actual needs, for example, W bins. Thus, the second exposure bin threshold point of each target object bin interval can be determined according to the second total exposure data P and the second preset number of bins, that is, ((P / W), 2*(P / W), ..., W*(P / W)). The second exposure bin threshold point is the bin critical point for equal-frequency binning of the target object data according to the angle of the exposure data.

[0194] In this way, the target object data are sorted in order from small to large, and the second accumulated exposure data of the target object behind the target object data are accumulated in order from small to large. When the second accumulated exposure data reaches (P / W), the target object data of the corresponding target object a is the target object binning interval threshold of the first bin, that is, the target objects whose target object data is less than or equal to the target object data of target object a are all classified into the first bin.

[0195] Correspondingly, when the second accumulated exposure data reaches (2*(P / W)), the target object data of the corresponding target object b is the target object binning interval threshold of the second bin, that is, the target object data is greater than the target object data of the target object a and less than or equal to the target object data of the target object b, and is classified into the second bin. Similarly, each target object data is classified into the corresponding bin, and equal-frequency binning processing is performed on the target object data according to the second exposure binning threshold point. This processing method performs equal-frequency binning from the perspective of exposure data, which can make the data covered in each bin after the equal-frequency binning processing more balanced, thereby improving the rationality of subsequent object quality data setting.

[0196] In some implementations, performing equal-frequency binning processing on the target object data based on the second exposure binning threshold point includes:

[0197] (2.1) sorting the target object data in ascending order;

[0198] (2.2) accumulating the second exposure data associated with the sorted target object data one by one according to the sorting, to obtain the second accumulated exposure data corresponding to each sorted target object data;

[0199] (2.3) using the target object data corresponding to the second accumulated exposure data that first reaches the second exposure binning threshold point as the target object binning interval threshold, and determining the target object binning interval according to the target object binning interval threshold;

[0200] (2.4) The target object data is binned based on the target object binning interval.

[0201] Among them, the target object data can be sorted in ascending order first. For example, there are 6 target objects, the target object data of target object 1 is 100, the target object data of target object 2 is 50, the target object data of target object 3 is 120, the target object data of target object 4 is 200, the target object data of target object 5 is 80, and the target object data of target object 6 is 300.

[0202] The target object data are sorted in ascending order, that is, the target object data of target object 2 is 50, the target object data of target object 5 is 80, the target object data of target object 1 is 100, the target object data of target object 3 is 120, the target object data of target object 4 is 200 and the target object data of target object 6 is 300.

[0203] Each target object includes both target object data and second exposure data, so that the target object data of each target object can be associated with the corresponding second exposure data. For example, the second exposure data of target object 1 is 50, the second exposure data of target object 2 is 10, the second exposure data of target object 3 is 40, the second exposure data of target object 4 is 60, the second exposure data of target object 5 is 20, and the second exposure data of target object 6 is 60. The second exposure data of all target objects are counted to obtain a second total exposure data of 240. In the embodiment of the present application, the second preset number of bins is 3, so the second exposure bin threshold point is (80, 160 and 240). The second exposure bin threshold point is the critical point for equal-frequency binning of the target object data according to the angle of the exposure data.

[0204] Furthermore, the second exposure data associated with the sorted target object data can be accumulated one by one in the sorting order to obtain the second accumulated exposure data corresponding to each sorted target object data, for example, the second accumulated exposure data of the sorted target object 2 is 10, the second accumulated exposure data of the target object 5 is 30, the second accumulated exposure data of the target object 1 is 80, the second accumulated exposure data of the target object 3 is 120, the second accumulated exposure data of the target object 4 is 180, and the second accumulated exposure data of the target object 6 is 240.

[0205] Thus, the target object data 100 of the target object 1 that first reaches the second exposure binning threshold point 80 is used as the target object binning interval threshold of the first bin; the target object data 200 of the target object 4 that first reaches the second exposure binning threshold point 160 is used as the target object binning interval threshold of the second bin; the target object data 300 of the target object 6 that first reaches the second exposure binning threshold point 240 is used as the target object binning interval threshold of the third bin. Accordingly, the target object binning intervals ((0-100], (100-200], (200-300]) can be determined according to the target object binning interval thresholds 100, 200, and 300.

[0206] Based on the target object binning interval, target object 2, target object 5 and target object 1 whose target object data is less than 100 are all classified into the first bin; target object 3 and target object 4 whose target object data is greater than 100 and less than or equal to 200 are both classified into the second bin; target object 6 whose target object data is greater than 200 and less than or equal to 300 is classified into the third bin, so as to realize binning processing of target object data, so that the data covered in each bin after equal-frequency binning processing is better balanced.

[0207] In step 203, the promotional position value data and the object quality data are combined to obtain target combination data.

[0208] In the related art, the allocation of computing resources only based on the promotion position value will result in too few computing resources being allocated to target objects with high quality in promotion positions with low promotion position values, which will lead to poor delivery effects, dissatisfaction of target objects with high quality in delivery effects, and even loss of target objects with high quality. Correspondingly, it will also result in too many computing resources being allocated to target objects with low quality in promotion positions with high promotion position values. Since the demand and consumption capacity of target objects with low quality for products are limited, it will lead to a waste of computing resources. Therefore, the above-mentioned method of allocating computing resources based on the characteristics of single-dimensional data will make the allocation of computing resources unreasonable, resulting in a waste of computing resources.

[0209] The embodiments of the present application can solve the above-mentioned technical problems and realize a multi-dimensional computing resource allocation method, combining the promotion position value data representing the promotion position and the object quality data representing the target object to obtain target combination data after the combination. In one embodiment, each promotion position value data and each object quality data can be added to obtain the target combination data under each <promotion position, target object>, and the target combination data represents the promotion value under the corresponding <promotion position, target object> combination. The higher the value of the target combination data, the higher the promotion value of the corresponding <promotion position, target object>, and more computing resources can be appropriately allocated. The lower the value of the target combination data, the lower the promotion value of the corresponding <promotion position, target object>, and the allocated computing resources can be appropriately reduced.

[0210] It should be noted that, due to the large differences in the promotion position data of the promotion position and the large differences in the target object data of the target object, the two are not combinable. In order to overcome this problem, the embodiments of the present application perform box processing on the promotion position data of the promotion position and the target object data of the target object to obtain promotion position value data representing different promotion position value levels and object quality data representing different target object quality levels, so that the two are combinable and can be combined to obtain the corresponding <promotion position, target object> target combination data.

[0211] In some implementations, the proportion of the promotional position value data and the object quality data in the target combination data can be adjusted by weight adjustment. In this embodiment, Figure 5 As shown, step 203 may include:

[0212] In step 301, the promotion position value data is combined with the first weight to obtain target promotion position value data.

[0213] In step 302, the object quality data is combined with the second weight to obtain target object quality data.

[0214] In step 303, the target promotion position value data and the target object quality data are summed to obtain target combination data.

[0215] Among them, the values ​​of the first weight and the second weight are 0 to 1, and the sum of the two is 1. The promotion position value data can be multiplied by the first weight to obtain the target promotion position value data, and the object quality data can be multiplied by the second weight to obtain the target object quality data. Finally, the target promotion position value data and the target object quality data are summed to obtain the target combination data.

[0216] The first weight and the second weight are used to adjust the proportion of the promotion position value data and the object quality data in the target combination data. Assuming that the first weight is greater than the second weight, for example, the first weight is 0.7 and the second weight is 0.3, then the proportion of the promotion position value data in the target combination data is higher than the object quality data. Assuming that the first weight is equal to the second weight, for example, the first weight is 0.5 and the second weight is 0.5, then the proportion of the promotion position value data in the target combination data is equal to the object quality data. Assuming that the first weight is less than the second weight, for example, the first weight is 0.3 and the second weight is 0.7, then the proportion of the promotion position value data in the target combination data is lower than the object quality data.

[0217] In some embodiments, the access traffic of the promotion position is different at different times, resulting in different corresponding background load rates of the promotion position. For example, in the late night time period, the access traffic of the promotion position is relatively small, and the corresponding background load rate of the promotion position is relatively small. In order to make full use of computing resources, more computing resources can be allocated for calculation. In the evening time period, the access traffic of the promotion position is relatively large, and the corresponding background load rate of the promotion position is large. In order to avoid excessive load and resulting in downtime, the allocation of computing resources can be appropriately reduced. Therefore, the corresponding background load rate of the promotion position can also be taken into consideration in the process of allocating computing resources, that is, the embodiment of the present application can also include: obtaining the target background load rate corresponding to the promotion position in the preset time period.

[0218] Among them, the preset time period can be any time period close to the current time. For example, if the current time is 22:01, the preset time period can be the time period from 21:56 to 22:00 or the time period from 21:01 to 22:00. Since the background load rate fluctuates, the background load rate every minute may be different. Therefore, the target background load rate can be the average background load rate of the promotion position in the preset time period. In this way, the target background load rate corresponding to each promotion position in the preset time period can be obtained. Since each promotion position has different number of users, the corresponding target background load rate is also different.

[0219] In some implementations, obtaining the target background load rate corresponding to the promotion position in a preset time period includes:

[0220] (1) obtaining a first background load rate of the promotion position in a preset time period and a second background load rate in a historical preset time period;

[0221] (2) combining the first background load rate with the third weight to obtain a first target background load rate;

[0222] (3) combining the second background load rate with the fourth weight to obtain a second target background load rate;

[0223] (4) The first target background load rate and the second target background load rate are summed to obtain the target background load rate of the promotion position.

[0224] Among them, the background load rate of the promotion position has a certain regularity in the same time period every day. Therefore, the embodiment of the present application can not only obtain the first background load rate of the promotion position in the preset time period, but also obtain the second background load rate in the historical preset time period. For example, the current time is 22:01, the preset time period can be the time period from 21:56 to 22:00 on the same day, and the historical preset time period can be the time period from 21:56 to 22:00 yesterday or the time period from 21:56 to 22:00 on the same day last week. That is, the embodiment of the present application can obtain the first background load rate of the current time period from 21:56 to 22:00, and can also obtain the second background load rate of the time period from 21:56 to 22:00 yesterday.

[0225] Furthermore, the values ​​of the third weight and the fourth weight are between 0 and 1, and the sum of the two is 1. The first background load rate can be combined with the third weight to obtain the first target background load rate, and the second background load rate can be combined with the fourth weight to obtain the second target background load rate. Finally, the first target background load rate and the second target background load rate are summed to obtain the target background load rate of the promotion position. Since the target background load rate is combined with the historical load rate in the same time period, the target background load rate is more in line with actual usage patterns, making the target background load rate more accurate.

[0226] The third weight and the fourth weight are used to adjust the proportion of the first background load rate and the second background load rate in the target background load rate. Assuming that the third weight is greater than the fourth weight, for example, the third weight is 0.8 and the second weight is 0.2, then the proportion and importance of the first background load rate in the target background load rate are higher than those of the second background load rate. Assuming that the third weight is equal to the fourth weight, for example, the third weight is 0.5 and the fourth weight is 0.5, then the proportion and importance of the first background load rate in the target background load rate are equal to the second background load rate. Assuming that the third weight is less than the fourth weight, for example, the first weight is 0.2 and the second weight is 0.8, then the proportion and importance of the first background load rate in the target background load rate are less than the second background load rate.

[0227] In some embodiments, the historical preset time period may be a preset time period of the previous day and a preset time period of the same day last week, thereby obtaining a first background load rate t1 of the promotion position in the preset time period, obtaining a first sub-background load rate t2 of the promotion position in the preset time period of the previous day, and obtaining a second sub-background load rate t3 of the promotion position in the preset time period of the same day last week. The first sub-background load rate t2 and the second sub-background load rate t3 constitute a second background load rate, and the target background load rate is calculated with reference to the following formula:

[0228] t=b*t1+c*t2+(1-bc)*t3

[0229] Among them, t is the target background load rate, b is the third weight, c is the first sub-weight, (1-bc) is the second sub-weight, the first sub-weight and the second sub-weight constitute the fourth weight, and thus, the product of the third weight and the first background load rate is calculated to obtain the first target background load rate, the first product of the first sub-weight and the first sub-background load rate is calculated, the first product is added to the second product of the second sub-weight and the second sub-background load rate to obtain the second target background load rate, and finally, the first target background load rate is added to the second target background load rate to obtain the target background load rate.

[0230] In some implementations, the value of the target combination data may be adjusted in combination with the busy / idle state of the target background load rate in the current time period. In this embodiment, Figure 6 As shown, step 203 may also include:

[0231] In step 402, the promotional position value data and the object quality data are summed to obtain intermediate combined data.

[0232] In step 402, the ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

[0233] Among them, the promotion position value data and the object quality data can be summed to obtain the intermediate combination data of each <promotion position, target object>. In order to combine the busy and idle status of the target background load rate in the current time period, the ratio of the intermediate combination data of each <promotion position, target object> and the corresponding target background load rate can be calculated to obtain the target combination data.

[0234] It can be seen that, assuming that the target background load rate is small, it means that the background computing burden of the promotion position is small. The target combination data calculated by the ratio of the intermediate combination data and the target background load rate will be larger, which means that more computing resources can be allocated. On the contrary, assuming that the target background load rate is large, it means that the background computing burden of the promotion position is large. The target combination data calculated by the ratio of the intermediate combination data and the target background load rate will be smaller, which means that some computing resources can be appropriately reduced. Therefore, after combining the target background load rate, the target combination data can also be adaptively adjusted according to the busy and idle state of the promotion position background load, so as to better balance the background computing resources and avoid waste of background computing resources and downtime caused by overload computing.

[0235] In step 204, binning is performed according to the target combination data to obtain binned target value data.

[0236] Among them, the target combination data can represent the push value of the corresponding <promotion position, target object>, but because the target combination data is combined through multi-dimensional data, the differences between the target combination data are relatively large. In order to realize the subsequent allocation of computing resources, it is necessary to process the target combination data in batches. In the embodiment of the present application, the target combination number can be binned by binning, that is, each target combination data can be sorted in order from small to large, and the number of bins can be determined according to the number of bins required. The number of bins can be set according to actual needs, and no specific quantity limit is made here.

[0237] In one embodiment, the target combination data may be binned using the following formula:

[0238] z=bin(x+y)

[0239] Here z is the target value data, bin() is the binning function, x is the promotion position value data, y is the object quality data, and x+y is the target combination data.

[0240] In one embodiment, the target combination data may also be binned using the following formula:

[0241] z=bin(a*x+(1-a)y)

[0242] Here, z is the target value data, a is the first weight, b is the second weight, x is the promotion position value data, y is the object quality data, and a*x+(1-a)y is the target combination data.

[0243] In one embodiment, the target combination data may also be binned using the following formula:

[0244] z = bin((x+y) / t)

[0245] Here z is the target value data, bin() is the binning function, x is the promotion position value data, y is the object quality data, t is the target background load rate, and (x+y) / t is the target combination data.

[0246] In one embodiment, the target combination data may also be binned using the following formula:

[0247] z=bin((a*x+(1-a)y) / t)

[0248] Here z is the target value data, a is the first weight, b is the second weight, x is the promotion position value data, y is the object quality data, t is the target background load rate, and (a*x+(1-a)y) / t is the target combination data.

[0249] In this way, the sorted target combination data are binned in turn according to the number of bins, and each target combination data is binned into a corresponding interval. The binning method can be equal-frequency binning or equal-distance binning to achieve binning of all target combination data, and each interval can correspond to a target value data. In this way, the target value data of each <promotion position, target object> after binning is obtained. The higher the target value data, the higher the promotion value of the corresponding <promotion position, target object>, and more computing resources can be appropriately allocated. The lower the target value data, the lower the promotion value of the corresponding <promotion position, target object>, and the allocated computing resources can be appropriately reduced.

[0250] For example, the number of bins is 5, which means there are 5 target bin intervals, bin 1 corresponds to target value data -2, bin 2 corresponds to target value data -1, bin 3 corresponds to target value data 0, bin 4 corresponds to target value data 1, and bin 5 corresponds to target value data 2. Based on this, all target combination data can be divided into 5 parts according to the sorting, and assigned to the corresponding target bin intervals to realize binning processing. According to the result of the binning processing, the target value data of each target combination data after binning is obtained, that is, the target value data of each <promotion position, target object> can be obtained. The target value data combines the characteristics of the promotion position value data of the promotion position and the characteristics of the object quality data of the target object, which can make the subsequent allocation of computing resources more reasonable and avoid waste of computing resources.

[0251] In one embodiment, the target value data can not only integrate the characteristics of the promotion position value data of the promotion position and the characteristics of the object quality data of the target object, but also combine the characteristics of the busy and idle status of the promotion position background load, so as to make the subsequent allocation of computing resources more reasonable and avoid the waste of computing resources. It can also better balance the background computing resources and avoid the waste of background computing resources and the situation where the background computing resources are down due to overload computing.

[0252] In some implementations, the target value data obtained by binning the target combination data includes:

[0253] (1) binning the target combined data;

[0254] (2) Determine the target value data of the target combination data based on the target binning interval in which the target combination data is located.

[0255] Among them, the embodiment of the present application can perform equal-width binning processing on the target combination data, so as to realize the grading processing of target combination data with different differences. The number of bins can be set according to actual needs. For example, 5 bins means that there are 5 target bin intervals, bin 1 corresponds to target value data -2, bin 2 corresponds to target value data -1, bin 3 corresponds to target value data 0, bin 4 corresponds to target value data 1, and bin 5 corresponds to target value data 2.

[0256] For example, there are 8 target combination data, the target combination data corresponding to <promotion position, target object>1 is -1, the target combination data corresponding to <promotion position, target object>2 is 2, the target combination data corresponding to <promotion position, target object>3 is 0, the target combination data corresponding to <promotion position, target object>4 is -1, the target combination data corresponding to <promotion position, target object>5 is -2, the target combination data corresponding to <promotion position, target object>6 is 1, the target combination data corresponding to <promotion position, target object>7 is 2, and the target combination data corresponding to <promotion position, target object>8 is 0.

[0257] Therefore, first sort the 8 <promotion position, target object> according to the target combination data to obtain the sorted target combination data, that is, the target combination data corresponding to <promotion position, target object> 5 is -2, the target combination data corresponding to <promotion position, target object> 1 is -1, the target combination data corresponding to <promotion position, target object> 4 is -1, the target combination data corresponding to <promotion position, target object> 3 is 0, the target combination data corresponding to <promotion position, target object> 8 is 0, the target combination data corresponding to <promotion position, target object> 6 is 1, the target combination data corresponding to <promotion position, target object> 2 is 2, and the target combination data corresponding to <promotion position, target object> 7 is 2.

[0258] According to the distance of the target combination data (-2, 2), and the number of bins 5, it is divided into 5 target bin intervals, namely the first target bin interval [-2], the second target bin interval [-1], the third target bin interval [0], the fourth target bin interval [1] and the fifth target bin interval [2]. Assign each target combination data to the corresponding target binning interval, that is, the target combination data corresponding to <promotion position, target object>5 is -2 and is assigned to the first target binning interval, the target combination data corresponding to <promotion position, target object>1 is -1 and the target combination data corresponding to <promotion position, target object>4 is -1 and is assigned to the second target binning interval, the target combination data corresponding to <promotion position, target object>3 is 0 and the target combination data corresponding to <promotion position, target object>8 is 0 and is assigned to the third target binning interval, the target combination data corresponding to <promotion position, target object>6 is 1 and is assigned to the fourth target binning interval, the target combination data corresponding to <promotion position, target object>2 is 2 and the target combination data corresponding to <promotion position, target object>7 is 2 and is assigned to the fifth target binning interval, thereby realizing equal-width binning processing of the target combination data.

[0259] Based on this, we can obtain the target value data corresponding to each target combination data according to the target box interval in which the target combination data is located, that is, the target value data of each <promotion position, target object> can be obtained. For example, the target value data corresponding to <promotion position, target object> 5 is -2, the target value data corresponding to <promotion position, target object> 1 and <promotion position, target object> 4 are -1, the target value data corresponding to <promotion position, target object> 3 and <promotion position, target object> 8 are 0, the target value data corresponding to <promotion position, target object> 6 is 1, and the target value data corresponding to <promotion position, target object> 2 and <promotion position, target object> 7 are 2. That is to say, the target value data of <Promotion position, target object> 5 is the lowest level, the target value data of <Promotion position, target object> 1 and <Promotion position, target object> 4 are the second lowest level, the target value data of <Promotion position, target object> 3 and <Promotion position, target object> 8 are medium level, the target value data of <Promotion position, target object> 6 is the second highest level, and the target value data of <Promotion position, target object> 2 and <Promotion position, target object> 7 are the highest level.

[0260] In step 205, based on the target binning interval where the target value data is located, the adjustment weight of the target object under the promotion position is obtained, and the corresponding candidate queue length is determined based on the adjustment weight.

[0261] In the related art, the initial queue length of each promotion position in each stage is fixed, that is, the computing resources of each promotion position are initially the same. In order to achieve flexible computing resource allocation, the embodiment of the present application can set a corresponding adjustment weight for the target bin interval of each gear. The lower the target bin interval, the less computing resource allocation can be appropriately reduced. A smaller adjustment weight can be set to shorten the subsequent corresponding candidate queue length. The higher the target bin interval, the more computing resources can be allocated. A larger adjustment weight can be set to lengthen the corresponding candidate queue length.

[0262] For example, the number of bins is 5, with 5 target bin intervals, each target bin interval is a gear, and a corresponding adjustment weight can be set for each target bin interval. To better illustrate the embodiment of the present application, please refer to the following table:

[0263] Gear Target value data Adjust weight Highest gear (high value) 2 20% (i.e. 20% longer) Second highest position (second highest value) 1 10% (i.e. 10% longer) Medium range (medium value) 0 0 (no change) Second lowest gear (second lowest value) -1 -10% (10% shorter) Lowest gear (low value) 2 -20 (i.e. 20% shorter)

[0264] Based on the above table, the corresponding adjustment weight can be obtained based on the target binning interval of the target value data of each <promotion position, target object>, and the initial queue length can be processed according to the adjustment weight to determine the candidate queue length of each <promotion position, target object> after adjustment. It can be seen that for the <promotion position, target object> with high promotion value, the candidate queue length can be increased by 20%, and as the candidate queue length increases, the number of promotion information that needs to be screened will also increase, more computing resources need to be allocated, and the accuracy of the promotion information screened for delivery will be greatly improved; for the <promotion position, target object> with the second highest promotion value, the candidate queue length can be increased by 10%, and as the candidate queue length increases slightly, the number of promotion information that needs to be screened will also increase slightly, and the computing resources need to be allocated With the increase of computing resources, the accuracy of the promotion information finally screened out for delivery will also increase slightly; for the <promotion position, target object> with medium promotion value, the length of the candidate queue will not change, the corresponding number of promotion information that needs to be screened will not change, the computing resources that need to be allocated will remain unchanged, and the accuracy of the promotion information finally screened out for delivery will remain unchanged; for the <promotion position, target object> with the second lowest promotion value, the length of the candidate queue can be reduced by 10%, and as the length of the candidate queue decreases, the corresponding number of promotion information that needs to be screened will also decrease slightly, and the computing resources that need to be allocated will also decrease slightly; for the <promotion position, target object> with low promotion value, the length of the candidate queue can be reduced by 20%, and as the length of the candidate queue decreases, the corresponding number of promotion information that needs to be screened will also decrease, and the computing resources that need to be allocated will decrease.

[0265] In this way, flexible allocation of computing resources is achieved, allocating more computing resources to <promotion position, target object> with higher promotion value, and reducing the allocation of computing resources to <promotion position, target object> with lower promotion value.

[0266] Moreover, since the target value data integrates the characteristics of the promotion position value data of the promotion position and the characteristics of the object quality data source of the target object, the promotion position value characteristics of the promotion position and the object quality characteristics of the target object are considered in multiple dimensions, making it more reasonable to allocate computing resources based on the target value data and avoiding waste of computing resources.

[0267] In one embodiment, the target value data can not only integrate the characteristics of the promotion position value data of the promotion position and the characteristics of the object quality data of the target object, but also combine the characteristics of the busy and idle status of the promotion position background load, so that the allocation of computing resources based on the target value data will be more reasonable, avoiding the waste of computing resources, and better balancing the background computing resources, avoiding the waste of background computing resources and the situation where the background is down due to overload computing.

[0268] In some implementations, determining the corresponding candidate queue length based on the adjustment weight includes:

[0269] (1) Obtain the initial queue length for the promotion position;

[0270] (2) Calculate the product of the initial queue length and the adjustment weight to determine the candidate queue length of the target object under the promotion position.

[0271] Among them, the initial queue length under each promotion position can be obtained, and the initial queue length includes the length of the candidate queue in the rough sorting stage and the length of the candidate queue in the fine sorting stage. The initial queue length is multiplied by the adjustment weight of the target object under the promotion position to determine the corresponding candidate queue length of the target object under the promotion position. For the <promotion position, target object> with higher promotion value, the candidate queue length is increased and more computing resources are allocated. For the <promotion position, target object> with lower promotion value, the candidate queue length is shortened and the allocation of computing resources is reduced, so as to realize flexible adjustment of the candidate queue length and further realize flexible adjustment of computing resources.

[0272] For example, the target value data corresponding to <Promotion Position, Target Object>5 is -2, the target value data corresponding to <Promotion Position, Target Object>1 and <Promotion Position, Target Object>4 is -1, the target value data corresponding to <Promotion Position, Target Object>3 and <Promotion Position, Target Object>8 is 0, the target value data corresponding to <Promotion Position, Target Object>6 is 1, and the target value data corresponding to <Promotion Position, Target Object>2 and <Promotion Position, Target Object>7 is 2. Then the candidate queue length corresponding to <Promotion Position, Target Object>5 will be shortened by 20%, the candidate queue length corresponding to <Promotion Position, Target Object>1 and <Promotion Position, Target Object>4 will be shortened by 10%, the candidate queue length corresponding to <Promotion Position, Target Object>3 and <Promotion Position, Target Object>8 will remain unchanged, the candidate queue length corresponding to <Promotion Position, Target Object>6 will be lengthened by 10%, and the candidate queue lengths of <Promotion Position, Target Object>2 and <Promotion Position, Target Object>7 will be lengthened by 20%.

[0273] As can be seen from the above, the embodiment of the present application obtains the promotion position data of the promotion position, and performs binning processing according to the promotion position data to obtain the promotion position value data after binning processing; obtains the target object data of the target object, and performs binning processing according to the target object data to obtain the object quality data after binning processing; combines the promotion position value data and the object quality data to obtain the target combination data; performs binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, obtains the adjustment weight of the target object under the promotion position, and determines the corresponding candidate queue length based on the adjustment weight. In this way, the promotion position value data of the promotion position and the object quality data of the target object are combined to process the candidate queue length, so as to realize the allocation of computing resources taking into account the data characteristics of the promotion position and the data characteristics of the target object at the same time. Compared with the scheme in the related art that only sets computing resources according to the promotion position value of the promotion position, the embodiment of the present application greatly improves the accuracy of computing resource allocation, thereby improving the utilization rate of computing resources.

[0274] In combination with the method described in the above embodiment, the method will be further described in detail with examples below.

[0275] In this embodiment, the queue processing device is specifically integrated into a server as an example for description.

[0276] Please also read Figure 7 , Figure 7 A schematic diagram of a process flow of a queue processing method provided in an embodiment of the present application. The process flow of the method may include:

[0277] In step 501, the server obtains consumption data and the number of requests for the promotional position, and calculates the corresponding cost per thousand requests data for the promotional position based on the consumption data and the number of requests for the promotional position.

[0278] Among them, the server can offline count the consumption data and promotion position request volume of each promotion position within a week, according to the formula: CPR (cost per thousand requests) = consumption data / promotion position request volume * 1000, divide the consumption data by the promotion position request volume and multiply by 1000 to get the corresponding cost per thousand requests data for each promotion position.

[0279] In step 502, the server uses the thousand-times-request cost data as promotional position data, obtains first exposure data associated with the promotional position data, and performs statistics on the first exposure data to obtain first total exposure data, and determines a first exposure bin threshold point based on the first total exposure data and a first preset bin number.

[0280] Among them, the server can use the thousand-times request cost data as promotion position data, and comprehensively calculate the relationship between resources and costs. Each promotion position also includes first exposure data. Therefore, each promotion position also includes promotion position data and first exposure data. In this way, the promotion position data of each promotion position can be associated with the corresponding first exposure data. It should be noted that the first exposure data on different promotion positions are different. In order to achieve the balance of exposure data covered in each promotion position bin interval after equal-frequency binning processing, the embodiment of the present application can perform equal-frequency binning processing from the perspective of exposure data.

[0281] In this way, the first exposure data corresponding to the promotion position behind each promotion position data can be obtained, and the first exposure data of all promotion positions are counted to obtain the first total exposure data of all promotion positions. For example, there are M promotion positions, each of which contains a number of first exposure data. The first exposure data of the M promotion positions are counted to obtain the first total exposure data N of the M promotion positions. M and N can be positive integers.

[0282] Furthermore, the first preset bin number is the bin number, which can be set according to actual needs, for example, X bins. Thus, the first exposure bin threshold point of each promotion position bin interval can be determined according to the first total exposure data N and the first preset bin number, that is, ((N / X), 2*(N / X), ..., X*(N / X)). The first exposure bin threshold point is the bin critical point for equal-frequency binning of the promotion position data according to the angle of the exposure data.

[0283] In step 503, the server sorts the promotion position data in ascending order, and accumulates the first exposure data associated with the sorted promotion position data one by one according to the sorting to obtain the first accumulated exposure data corresponding to each sorted promotion position data, and uses the promotion position data corresponding to the first accumulated exposure data that first reaches the first exposure bin threshold point as the promotion position bin interval threshold, and determines the promotion position bin interval according to the promotion position bin interval threshold, and bins the promotion position data based on the promotion position bin interval, and determines the promotion position value data of the promotion position according to the promotion position bin interval in which the promotion position data is located.

[0284] Among them, the server can first sort the promotion position data in ascending order. For example, there are 9 promotion positions, the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700, the promotion position data of promotion position 8 is 80, and the promotion position data of promotion position 9 is 900. The promotion position data are sorted in ascending order, that is, the promotion position data of promotion position 8 is 80, the promotion position data of promotion position 2 is 150, the promotion position data of promotion position 5 is 180, the promotion position data of promotion position 1 is 200, the promotion position data of promotion position 3 is 230, the promotion position data of promotion position 4 is 400, the promotion position data of promotion position 6 is 550, the promotion position data of promotion position 7 is 700 and the promotion position data of promotion position 9 is 900 in this order.

[0285] Each promotion position includes both promotion position data and first exposure data, so that the promotion position data of each promotion position can be associated with the corresponding first exposure data. For example, the first exposure data of promotion position 1 is 40, the first exposure data of promotion position 2 is 30, the first exposure data of promotion position 3 is 20, the first exposure data of promotion position 4 is 40, the first exposure data of promotion position 5 is 50, the first exposure data of promotion position 6 is 40, the first exposure data of promotion position 7 is 40, the first exposure data of promotion position 8 is 10, and the first exposure data of promotion position 9 is 60. The first exposure data of all promotion positions are counted to obtain the first total exposure data of 330. In the embodiment of the present application, the number of the first preset bins is 3, so the first exposure bin threshold point is (110, 220 and 330). The first exposure bin threshold point is the critical point for equal-frequency binning of the promotion position data according to the angle of the exposure data.

[0286] Furthermore, the first exposure data associated with the sorted promotion position data can be accumulated one by one in the sorting order to obtain the first accumulated exposure data corresponding to each sorted promotion position data. For example, the first accumulated exposure data of the sorted promotion position 8 is 10, the first accumulated exposure data of the promotion position 2 is 40, the first accumulated exposure data of the promotion position 5 is 90, the first accumulated exposure data of the promotion position 1 is 130, the first accumulated exposure data of the promotion position 3 is 150, the first accumulated exposure data of the promotion position 4 is 190, the first accumulated exposure data of the promotion position 6 is 230, the first accumulated exposure data of the promotion position 7 is 270 and the first accumulated exposure data of the promotion position 9 is 330.

[0287] Thus, the promotion position data 200 of promotion position 1 that first reaches the first exposure bin threshold point 110 is used as the promotion position bin interval threshold of the first bin; the promotion position data 550 of promotion position 6 that first reaches the first exposure bin threshold point 220 is used as the promotion position bin interval threshold of the second bin; the promotion position data 900 of promotion position 9 that first reaches the first exposure bin threshold point 330 is used as the promotion position bin interval threshold of the third bin. Accordingly, the promotion position bin intervals ((0-200], (200-550], (550-900]) can be determined according to the promotion position bin interval thresholds 200, 550 and 900.

[0288] Based on the promotion position bin interval, promotion position 8, promotion position 2, promotion position 5 and promotion position 1 with promotion position data less than 200 are all classified into the first bin; promotion position 3, promotion position 4 and promotion position 6 with promotion position data greater than 200 and less than or equal to 550 are all classified into the second bin; promotion position 7 and promotion position 9 with promotion position data greater than 550 and less than or equal to 900 are all classified into the third bin.

[0289] Based on this, the promotion position value data corresponding to each promotion position data can be obtained according to the promotion position bin interval in which the promotion position data is located, that is, the promotion position value data of each promotion position can be obtained. For example, the promotion position value data of promotion position 8, promotion position 2, promotion position 5 and promotion position 1 is -1, the promotion position value data of promotion position 3, promotion position 4 and promotion position 6 is 0, and the promotion position value data of promotion position 7 and promotion position 9 is 1. That is to say, the promotion position value data of promotion position 8, promotion position 2, promotion position 5 and promotion position 1 are the lowest level, the promotion position value data of promotion position 3, promotion position 4 and promotion position 6 are the middle level, and the promotion position value is higher than the lowest level, and the promotion position value data of promotion position 7 and promotion position 9 are the highest level, and the promotion position value is higher than the middle level. In this way, the binning processing of the promotion position data according to the exposure data is realized, so that the data covered in each bin after the equal-frequency binning processing is better balanced.

[0290] In step 504, the server obtains the cost per thousand exposures data of the target object, uses the cost per thousand exposures data as the target object data, obtains second exposure data associated with the target object data, and performs statistics on the second exposure data to obtain second total exposure data, and determines a second exposure bin threshold point based on the second total exposure data and the second preset bin number.

[0291] The server may obtain the average cost per thousand exposures of the target object's most recent 50 promotional information exposures within a month, and use the average as the cost per thousand exposures of the target object to ensure the real-time cost per thousand exposures of each target object.

[0292] In this way, the thousand-times-exposure cost data is used as the target object data, and each target object also includes second exposure data, which refers to the number of exposures of the promotional information exposed on the target object. Therefore, each target object includes both the target object data and the second exposure data, and thus, the target object data of each target object can be associated with the corresponding second exposure data. It should be noted that the second exposure data on different target objects is different. In order to achieve the balance of the exposure data covered in each target object binning interval after the equal-frequency binning processing, the embodiment of the present application can perform equal-frequency binning processing from the perspective of the exposure data.

[0293] In this way, the second exposure data corresponding to the target object behind each target object data can be obtained, and the second exposure data of all target objects are counted to obtain the second total exposure data of all target objects. For example, there are Y target objects, each of which contains a number of second exposure data. The second exposure data of the Y target objects are counted to obtain the second total exposure data P of the Y target objects. Y and P can be positive integers.

[0294] Furthermore, the second preset number of bins is the number of bins, which can be set according to actual needs, for example, W bins. Thus, the second exposure bin threshold point of each target object bin interval can be determined according to the second total exposure data P and the second preset number of bins, that is, ((P / W), 2*(P / W), ..., W*(P / W)). The second exposure bin threshold point is the bin critical point for equal-frequency binning of the target object data according to the angle of the exposure data.

[0295] In step 505, the server sorts the target object data in ascending order, accumulates the second exposure data associated with the sorted target object data one by one according to the sorting, to obtain the second accumulated exposure data corresponding to each sorted target object data, and uses the target object data corresponding to the second accumulated exposure data that first reaches the second exposure binning threshold point as the target object binning interval threshold, determines the target object binning interval according to the target object binning interval threshold, performs binning processing on the target object data based on the target object binning interval, and determines the object quality data of the target object according to the target object binning interval in which the target object data is located.

[0296] Among them, the target object data can be sorted in ascending order first. For example, there are 6 target objects, the target object data of target object 1 is 100, the target object data of target object 2 is 50, the target object data of target object 3 is 120, the target object data of target object 4 is 200, the target object data of target object 5 is 80, and the target object data of target object 6 is 300.

[0297] The target object data are sorted in ascending order, that is, the target object data of target object 2 is 50, the target object data of target object 5 is 80, the target object data of target object 1 is 100, the target object data of target object 3 is 120, the target object data of target object 4 is 200 and the target object data of target object 6 is 300.

[0298] Each target object includes both target object data and second exposure data, so that the target object data of each target object can be associated with the corresponding second exposure data. For example, the second exposure data of target object 1 is 50, the second exposure data of target object 2 is 10, the second exposure data of target object 3 is 40, the second exposure data of target object 4 is 60, the second exposure data of target object 5 is 20, and the second exposure data of target object 6 is 60. The second exposure data of all target objects are counted to obtain a second total exposure data of 240. In the embodiment of the present application, the second preset number of bins is 3, so the second exposure bin threshold point is (80, 160 and 240). The second exposure bin threshold point is the critical point for equal-frequency binning of the target object data according to the angle of the exposure data.

[0299] Furthermore, the second exposure data associated with the sorted target object data can be accumulated one by one in the sorting order to obtain the second accumulated exposure data corresponding to each sorted target object data, for example, the second accumulated exposure data of the sorted target object 2 is 10, the second accumulated exposure data of the target object 5 is 30, the second accumulated exposure data of the target object 1 is 80, the second accumulated exposure data of the target object 3 is 120, the second accumulated exposure data of the target object 4 is 180, and the second accumulated exposure data of the target object 6 is 240.

[0300] Thus, the target object data 100 of the target object 1 that first reaches the second exposure binning threshold point 80 is used as the target object binning interval threshold of the first bin; the target object data 200 of the target object 4 that first reaches the second exposure binning threshold point 160 is used as the target object binning interval threshold of the second bin; the target object data 300 of the target object 6 that first reaches the second exposure binning threshold point 240 is used as the target object binning interval threshold of the third bin. Accordingly, the target object binning intervals ((0-100], (100-200], (200-300]) can be determined according to the target object binning interval thresholds 100, 200, and 300.

[0301] Based on the target object binning interval, target object 2, target object 5 and target object 1 whose target object data is less than 100 are all classified into the first bin; target object 3 and target object 4 whose target object data is greater than 100 and less than or equal to 200 are both classified into the second bin; target object 6 whose target object data is greater than 200 and less than or equal to 300 is classified into the third bin, so as to realize binning processing of target object data, so that the data covered in each bin after equal-frequency binning processing is better balanced.

[0302] Based on this, the object quality data corresponding to each target object data can be obtained according to the target object binning interval in which the target object data is located, that is, the object quality data of each target object can be obtained. For example, the object quality data of target object 2, target object 5 and target object 1 is -1, the object quality data of target object 3 and target object 4 is 0, and the object quality data of target object 6 is 1. That is to say, the object quality data of target object 2, target object 5 and target object 1 are at the lowest level, the object quality data of target object 3 and target object 4 are at the middle level, and the object quality data are higher than the lowest level, and the object quality data of target object 6 is at the highest level, and the object quality data is higher than the middle level. In this way, the object quality data of different target objects can be binned according to the order of target object data from high to low.

[0303] In step 506, the server obtains the first background load rate of the promotion position in the preset time period and the second background load rate in the historical preset time period, combines the first background load rate with the third weight to obtain the first target background load rate, combines the second background load rate with the fourth weight to obtain the second target background load rate, and sums the first target background load rate and the second target background load rate to obtain the target background load rate of the promotion position.

[0304] The preset time period of the history may be the preset time period of the previous day and the preset time period of the same day last week, thereby obtaining the first background load rate t1 of the promotion position in the preset time period, obtaining the first sub-background load rate t2 of the promotion position in the preset time period of the previous day, and obtaining the second sub-background load rate t3 of the promotion position in the preset time period of the same day last week. The first sub-background load rate t2 and the second sub-background load rate t3 constitute the second background load rate, and the target background load rate is calculated with reference to the following formula:

[0305] t=b*t1+c*t2+(1-bc)*t3

[0306] Among them, t is the target background load rate, b is the third weight, c is the first sub-weight, (1-bc) is the second sub-weight, and the first sub-weight and the second sub-weight constitute the fourth weight. In this way, the product of the third weight and the first background load rate is calculated to obtain the first target background load rate, the first product of the first sub-weight and the first sub-background load rate is calculated, and the first product is added to the second product of the second sub-weight and the second sub-background load rate to obtain the second target background load rate. Finally, the first target background load rate is added to the second target background load rate to obtain the target background load rate. The target background load rate combines the load rate in the same historical time period, so that the target background load rate is more in line with the actual usage rules, and the target background load rate is more accurate.

[0307] In step 507, the server combines the promotion position value data with the fifth weight to obtain intermediate promotion position value data, combines the object quality data with the sixth weight to obtain intermediate object quality data, and sums the intermediate promotion position value data and the intermediate object quality data to obtain intermediate combined data.

[0308] Among them, the values ​​of the fifth weight and the sixth weight are 0 to 1, and the sum of the two is 1. The promotion position value data can be multiplied by the fifth weight to obtain intermediate promotion position value data, and the object quality data can be multiplied by the sixth weight to obtain intermediate object quality data. Finally, the intermediate promotion position value data and the intermediate object quality data are summed to obtain intermediate combined data.

[0309] The fifth weight and the sixth weight are used to adjust the proportion of the promotion position value data and the object quality data in the intermediate combination data. Assuming that the fifth weight is greater than the sixth weight, for example, the fifth weight is 0.7 and the sixth weight is 0.3, then the proportion of the promotion position value data in the intermediate combination data is higher than the object quality data. Assuming that the fifth weight is equal to the sixth weight, for example, the fifth weight is 0.5 and the sixth weight is 0.5, then the proportion of the promotion position value data in the intermediate combination data is equal to the object quality data. Assuming that the fifth weight is less than the sixth weight, for example, the fifth weight is 0.3 and the sixth weight is 0.7, then the proportion of the promotion position value data in the intermediate combination data is lower than the object quality data.

[0310] In step 508, the server calculates the ratio of the intermediate combination data to the target background load rate to obtain the target combination data.

[0311] In order to combine the busy and idle status of the target background load rate in the current time period, the ratio of the intermediate combination data of each <promotion position, target object> and the corresponding target background load rate can be calculated to obtain the target combination data.

[0312] It can be seen that, assuming that the target background load rate is small, it means that the background computing burden of the promotion position is small. The target combination data calculated by the ratio of the intermediate combination data and the target background load rate will be larger, which means that more computing resources can be allocated. On the contrary, assuming that the target background load rate is large, it means that the background computing burden of the promotion position is large. The target combination data calculated by the ratio of the intermediate combination data and the target background load rate will be smaller, which means that some computing resources can be appropriately reduced. Therefore, after combining the target background load rate, the target combination data can also be adaptively adjusted according to the busy and idle state of the promotion position background load, so as to better balance the background computing resources and avoid waste of background computing resources and downtime caused by overload computing.

[0313] In step 509, the server performs equal-frequency binning processing according to the target combination data, and determines the target value data of the target combination data based on the target binning interval in which the target combination data is located.

[0314] Among them, the embodiment of the present application can perform equal-width binning processing on the target combination data, so as to realize the grading processing of target combination data with different differences. The number of bins can be set according to actual needs. For example, 5 bins means that there are 5 target bin intervals, bin 1 corresponds to target value data -2, bin 2 corresponds to target value data -1, bin 3 corresponds to target value data 0, bin 4 corresponds to target value data 1, and bin 5 corresponds to target value data 2.

[0315] For example, there are 8 target combination data, the target combination data corresponding to <promotion position, target object>1 is -1, the target combination data corresponding to <promotion position, target object>2 is 2, the target combination data corresponding to <promotion position, target object>3 is 0, the target combination data corresponding to <promotion position, target object>4 is -1, the target combination data corresponding to <promotion position, target object>5 is -2, the target combination data corresponding to <promotion position, target object>6 is 1, the target combination data corresponding to <promotion position, target object>7 is 2, and the target combination data corresponding to <promotion position, target object>8 is 0.

[0316] Therefore, first sort the 8 <promotion position, target object> according to the target combination data to obtain the sorted target combination data, that is, the target combination data corresponding to <promotion position, target object> 5 is -2, the target combination data corresponding to <promotion position, target object> 1 is -1, the target combination data corresponding to <promotion position, target object> 4 is -1, the target combination data corresponding to <promotion position, target object> 3 is 0, the target combination data corresponding to <promotion position, target object> 8 is 0, the target combination data corresponding to <promotion position, target object> 6 is 1, the target combination data corresponding to <promotion position, target object> 2 is 2, and the target combination data corresponding to <promotion position, target object> 7 is 2.

[0317] According to the distance of the target combination data (-2, 2), and the number of bins 5, it is divided into 5 target bin intervals, namely the first target bin interval [-2], the second target bin interval [-1], the third target bin interval [0], the fourth target bin interval [1] and the fifth target bin interval [2]. Assign each target combination data to the corresponding target binning interval, that is, the target combination data corresponding to <promotion position, target object>5 is -2 and is assigned to the first target binning interval, the target combination data corresponding to <promotion position, target object>1 is -1 and the target combination data corresponding to <promotion position, target object>4 is -1 and is assigned to the second target binning interval, the target combination data corresponding to <promotion position, target object>3 is 0 and the target combination data corresponding to <promotion position, target object>8 is 0 and is assigned to the third target binning interval, the target combination data corresponding to <promotion position, target object>6 is 1 and is assigned to the fourth target binning interval, the target combination data corresponding to <promotion position, target object>2 is 2 and the target combination data corresponding to <promotion position, target object>7 is 2 and is assigned to the fifth target binning interval, thereby realizing equal-width binning processing of the target combination data.

[0318] Based on this, we can obtain the target value data corresponding to each target combination data according to the target box interval in which the target combination data is located, that is, the target value data of each <promotion position, target object> can be obtained. For example, the target value data corresponding to <promotion position, target object> 5 is -2, the target value data corresponding to <promotion position, target object> 1 and <promotion position, target object> 4 are -1, the target value data corresponding to <promotion position, target object> 3 and <promotion position, target object> 8 are 0, the target value data corresponding to <promotion position, target object> 6 is 1, and the target value data corresponding to <promotion position, target object> 2 and <promotion position, target object> 7 are 2. That is to say, the target value data of <Promotion position, target object> 5 is the lowest level, the target value data of <Promotion position, target object> 1 and <Promotion position, target object> 4 are the second lowest level, the target value data of <Promotion position, target object> 3 and <Promotion position, target object> 8 are medium level, the target value data of <Promotion position, target object> 6 is the second highest level, and the target value data of <Promotion position, target object> 2 and <Promotion position, target object> 7 are the highest level.

[0319] In step 510, the server obtains the adjustment weight of the target object under the promotion position based on the target bin interval where the target value data is located, obtains the initial queue length under the promotion position, calculates the product of the initial queue length and the adjustment weight, and determines the candidate queue length of the target object under the promotion position.

[0320] To better illustrate the embodiments of this application, please refer to the following table:

[0321]

[0322]

[0323] Among them, the server obtains the adjustment weight of the target object under the promotion position based on the target binning interval where the target value data is located. For example, <Promotion Position, Target Object> 5 obtains an adjustment weight of -20%, <Promotion Position, Target Object> 1 and <Promotion Position, Target Object> 4 obtain an adjustment weight of -10%, <Promotion Position, Target Object> 3 and <Promotion Position, Target Object> 8 obtain an adjustment weight of 0, <Promotion Position, Target Object> 6 obtains an adjustment weight of 10%, and <Promotion Position, Target Object> 2 and <Promotion Position, Target Object> 7 obtain an adjustment weight of 20%.

[0324] Furthermore, the initial queue length under each promotion position can be obtained, and the initial queue length includes the length of the candidate queue in the rough sorting stage and the length of the candidate queue in the fine sorting stage. The initial queue length is multiplied by the adjustment weight of the target object under the promotion position to determine the corresponding candidate queue length of the target object under the promotion position. In this way, the candidate queue length corresponding to <Promotion position, target object>5 will be shortened by 20%, the candidate queue length corresponding to <Promotion position, target object>1 and <Promotion position, target object>4 will be shortened by 10%, the candidate queue length corresponding to <Promotion position, target object>3 and <Promotion position, target object>8 will remain unchanged, the candidate queue length corresponding to <Promotion position, target object>6 will be lengthened by 10%, and the candidate queue lengths of <Promotion position, target object>2 and <Promotion position, target object>7 will be lengthened by 20%.

[0325] To better illustrate the embodiments of this application, please refer to Figure 8 It is understood that the server can receive requests for the delivery of promotional information to promotional positions from different terminals, and it is necessary to screen out the most suitable promotional information for each request from the background and deliver it to the corresponding terminal for display. Therefore, the embodiment of the present application can calculate the corresponding target value data of each <promotional position, target object> by combining the promotional position value data of the promotion position, the object quality data of the target object and the target background load rate of the promotion position background, and dynamically allocate queues according to the target value data under multi-dimensional combination, set the candidate queue length of each <promotional position, target object> in 5 modes, and screen the promotion information based on the candidate queue after the length setting. For requests for promotional information delivery to promotional slots with high promotional slot value and target objects with high object quality, more computing resources will be allocated according to the circumstances, the length of the candidate queue will be increased, and the accuracy of the promotional information delivery will be improved. For requests for promotional slots with low promotional slot value and target objects with low object quality, the allocation of computing resources will be reduced and the length of the candidate queue will be shortened, so that the allocation of computing resources will be more reasonable and the waste of computing resources will be avoided. It can also better balance the backend computing resources and avoid the waste of backend computing resources and the downtime caused by overload computing, thereby achieving a win-win situation of operational goals and delivery effects.

[0326] Calculation example: Assume that the candidate queue length of application 1 in the fine ranking stage of the promotion position is 200, and the candidate queue length of application 2 in the fine ranking stage of the promotion position is 150, and the two applications are deployed using the same model.

[0327] There are target objects A and B. If this solution is not adopted, during the fine sorting phase, the number of requests for calculation by application 1 of target objects A and B is the same, both 200 times; the number of requests for calculation by application 2 of target objects A and B is also the same, both 150 times.

[0328] After adopting this solution, the calculated CPR of application 1 is higher, that is, the advertising space value of application 1 is higher than that of application 2; the CPM of target object A is higher than that of B, that is, the object quality of target object A is higher than that of B; after binning calculation, the queue of target object A can be lengthened by a corresponding amount on application 1 and application 2, and the queue of target object B can be shortened by a corresponding amount on these two applications.

[0329] 1) At 6 a.m., the target background load rate is not very high (time value). The number of calculations of target object A on application 1 is limited to 250 times, that is, the candidate queue can be increased by 50, and the number of calculations on application 2 is limited to 180 times, that is, the candidate queue can also be increased by 30; the number of calculations of target object B on application 1 is limited to 180 times, that is, the candidate queue is reduced by 20, and the number of calculations on application 2 is limited to 120 times, that is, the candidate queue is reduced by 30;

[0330] 2) At 8 o'clock in the evening, the request peak period, that is, the target background load rate is very high (time value), the calculation number of target object A on application 1 is limited to 210 times, that is, 10 candidate queues can be added, and the calculation number on application 1 is limited to 155 times, that is, 5 candidate queues can be added; the calculation number of target object B on application 1 is limited to 170 times, that is, 30 candidate queues are reduced, and the calculation number on application 1 is limited to 110 times, that is, 40 candidate queues are reduced.

[0331] As can be seen from the above, the embodiment of the present application obtains the promotion position data of the promotion position, and performs binning processing according to the promotion position data to obtain the promotion position value data after binning processing; obtains the target object data of the target object, and performs binning processing according to the target object data to obtain the object quality data after binning processing; combines the promotion position value data and the object quality data to obtain the target combination data; performs binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, obtains the adjustment weight of the target object under the promotion position, and determines the corresponding candidate queue length based on the adjustment weight. In this way, the promotion position value data of the promotion position and the object quality data of the target object are combined to process the candidate queue length, so as to realize the allocation of computing resources taking into account the data characteristics of the promotion position and the data characteristics of the target object at the same time. Compared with the scheme in the related art that only sets computing resources according to the promotion position value of the promotion position, the embodiment of the present application greatly improves the accuracy of computing resource allocation, thereby improving the utilization rate of computing resources.

[0332] Furthermore, the embodiment of the present application also introduces a target background load rate to perform further flexible processing on computing resources, which can better balance the background computing resources and avoid waste of background computing resources and downtime caused by overload computing.

[0333] In order to better implement the queue processing method provided in the embodiment of the present application, the embodiment of the present application also provides a device based on the above queue processing method. The meanings of the terms are the same as those in the above queue processing method, and the specific implementation details can refer to the description in the method embodiment.

[0334] See also Fig. 9 , Fig. 9 A schematic diagram of the structure of a queue processing device provided in an embodiment of the present application, wherein the queue processing device may include a first binning processing unit 601, a second binning processing unit 602, a combining unit 603, a third binning processing unit 604 and a determining unit 605, etc.

[0335] The first box processing unit 601 is used to obtain promotion position data of the promotion position, and perform box processing according to the promotion position data to obtain promotion position value data after box processing.

[0336] In some implementations, the first binning processing unit 601 includes:

[0337] The acquisition subunit (unlabeled) is used to obtain consumption data of the promotional position and the amount of promotional position requests;

[0338] A calculation subunit (not identified), used to calculate the cost per thousand requests corresponding to the promotion position according to the consumption data and the promotion position request amount;

[0339] A binning processing subunit (not identified), used to use the thousand-times-request cost data as promotion position data, and perform equal-frequency binning processing according to the promotion position data;

[0340] A determination subunit (unlabeled) is used to determine the promotion position value data of the promotion position according to the promotion position binning interval in which the promotion position data is located.

[0341] In some embodiments, the binning processing subunit (not labeled) is used to:

[0342] The thousand request cost data is used as promotion position data;

[0343] Acquire first exposure data associated with the promotion position data, and perform statistics on the first exposure data to obtain first total exposure data;

[0344] Determine a first exposure bin threshold point according to the first total exposure data and a first preset bin number;

[0345] Based on the first exposure binning threshold point, equal-frequency binning is performed on the promotion position data.

[0346] In some embodiments, the binning processing subunit (not identified) is further configured to:

[0347] The thousand request cost data is used as promotion position data;

[0348] Acquire first exposure data associated with the promotion position data, and perform statistics on the first exposure data to obtain first total exposure data;

[0349] Determine a first exposure bin threshold point according to the first total exposure data and a first preset bin number;

[0350] Sort the promotion position data in ascending order;

[0351] Accumulate the first exposure data associated with the sorted promotion position data one by one according to the sorting, to obtain the first accumulated exposure data corresponding to each sorted promotion position data;

[0352] The promotion position data corresponding to the first accumulated exposure data that first reaches the first exposure bin threshold point is used as the promotion position bin interval threshold, and the promotion position bin interval is determined according to the promotion position bin interval threshold;

[0353] The promotion position data is binned based on the promotion position binning interval.

[0354] The second binning processing unit 602 is used to obtain target object data of the target object, and perform binning processing according to the target object data to obtain object quality data after binning processing.

[0355] In some implementations, the second binning processing unit 602 includes:

[0356] An acquisition subunit (unlabeled) is used to acquire cost per thousand impressions data of a target object;

[0357] A binning processing subunit (not identified), used to use the thousand exposure cost data as target object data, and perform equal frequency binning processing according to the target object data;

[0358] A determination subunit (not identified) is used to determine the object quality data of the target object according to the target object binning interval in which the target object data is located.

[0359] In some embodiments, the binning processing subunit (not identified) is used to:

[0360] The cost per thousand impressions data is used as target object data;

[0361] Acquire second exposure data associated with the target object data, and perform statistics on the second exposure data to obtain second total exposure data;

[0362] Determine a second exposure bin threshold point according to the second total exposure data and a second preset bin number;

[0363] Based on the second exposure binning threshold point, equal-frequency binning processing is performed on the target object data.

[0364] In some embodiments, the binning processing subunit (not identified) is further configured to:

[0365] The cost per thousand impressions data is used as target object data;

[0366] Acquire second exposure data associated with the target object data, and perform statistics on the second exposure data to obtain second total exposure data;

[0367] Determine a second exposure bin threshold point according to the second total exposure data and a second preset bin number;

[0368] Sort the target object data in ascending order;

[0369] Accumulating the second exposure data associated with the sorted target object data one by one according to the sorting, to obtain second accumulated exposure data corresponding to each sorted target object data;

[0370] The target object data corresponding to the second accumulated exposure data that first reaches the second exposure binning threshold point is used as the target object binning interval threshold, and the target object binning interval is determined according to the target object binning interval threshold;

[0371] The target object data is binned based on the target object binning interval.

[0372] The combining unit 603 is used to combine the promotional position value data and the object quality data to obtain target combined data.

[0373] In some embodiments, the combination unit 603 is used to:

[0374] Combining the promotion position value data with the first weight to obtain target promotion position value data;

[0375] Combining the object quality data with the second weight to obtain target object quality data;

[0376] The target promotion position value data and the target object quality data are summed to obtain target combination data.

[0377] In some embodiments, the apparatus further comprises:

[0378] A load rate acquisition unit (not marked), used to acquire a target background load rate corresponding to the promotion position in a preset time period;

[0379] The combination unit 603 is further used for:

[0380] The promotion position value data and the object quality data are summed to obtain intermediate combination data;

[0381] The ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

[0382] In some embodiments, the load rate acquisition unit (not labeled) is further used to:

[0383] Obtaining a first background load rate of the promotion position in a preset time period and a second background load rate in a historical preset time period;

[0384] Combining the first background load rate with the third weight to obtain a first target background load rate;

[0385] Combining the second background load rate with the fourth weight to obtain a second target background load rate;

[0386] The first target background load rate and the second target background load rate are summed to obtain the target background load rate of the promotion position.

[0387] In some embodiments, the combination unit 603 is further used to:

[0388] The promotion position value data is combined with the fifth weight to obtain the intermediate promotion position value data;

[0389] combining the object quality data and the sixth weight to obtain intermediate object quality data;

[0390] The intermediate promotion position value data and the intermediate object quality data are summed to obtain intermediate combination data;

[0391] The ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

[0392] The third binning processing unit 604 is used to perform binning processing according to the target combination data to obtain target value data after binning processing.

[0393] In some embodiments, the third binning processing unit 604 is used to:

[0394] Binning is performed according to the target combined data;

[0395] Based on the target binning interval in which the target combination data is located, target value data of the target combination data is determined.

[0396] The determination unit 605 is used to obtain the adjustment weight of the target object in the promotion position based on the target binning interval where the target value data is located, and determine the corresponding candidate queue length based on the adjustment weight.

[0397] In some embodiments, the determining unit 605 is configured to:

[0398] Based on the target binning interval where the target value data is located, obtaining the adjustment weight of the target object under the promotion position;

[0399] Get the initial queue length for the promotion position;

[0400] The product of the initial queue length and the adjustment weight is calculated to determine the candidate queue length of the target object under the promotion position.

[0401] The specific implementation of each of the above units can be found in the previous embodiments, which will not be described in detail here.

[0402] As can be seen from the above, the embodiment of the present application obtains the promotion position data of the promotion position through the first binning processing unit 601, and performs binning processing according to the promotion position data to obtain the promotion position value data after binning processing; the second binning processing unit 602 obtains the target object data of the target object, and performs binning processing according to the target object data to obtain the object quality data after binning processing; the combination unit 603 combines the promotion position value data and the object quality data to obtain the target combination data; the third binning processing unit 604 performs binning processing according to the target combination data to obtain the target value data after binning processing; the determination unit 605 obtains the adjustment weight of the target object under the promotion position based on the target binning interval where the target value data is located, and determines the corresponding candidate queue length based on the adjustment weight. In this way, the promotion position value data of the promotion position and the object quality data of the target object are combined to process the candidate queue length, so as to realize the allocation of computing resources taking into account the data characteristics of the promotion position and the data characteristics of the target object at the same time. Compared with the scheme in the related art that only sets computing resources by the promotion position value of the promotion position, the embodiment of the present application greatly improves the accuracy of computing resource allocation, thereby improving the utilization rate of computing resources.

[0403] The present application also provides a computer device, such as Fig.10 As shown, it shows a schematic diagram of the structure of the server involved in the embodiment of the present application, specifically:

[0404] The computer device may include one or more processing core processors 701, one or more computer readable storage media memories 702, a power supply 703, an input unit 704 and other components. Those skilled in the art will appreciate that Fig.10 The computer device structure shown in the figure does not constitute a limitation on the computer device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Among them:

[0405] The processor 701 is the control center of the computer device. It uses various interfaces and lines to connect various parts of the entire computer device. By running or executing software programs and / or modules stored in the memory 702 and calling data stored in the memory 702, it executes various functions of the computer device and processes data, thereby monitoring the computer device as a whole. Optionally, the processor 701 may include one or more processing cores; optionally, the processor 701 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 701.

[0406] The memory 702 can be used to store software programs and modules. The processor 701 executes various functional applications and data processing by running the software programs and modules stored in the memory 702. The memory 702 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the server, etc. In addition, the memory 702 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 702 may also include a memory controller to provide the processor 701 with access to the memory 702.

[0407] The computer device also includes a power supply 703 for supplying power to various components. Optionally, the power supply 703 can be logically connected to the processor 701 through a power management system, so as to manage charging, discharging, and power consumption through the power management system. The power supply 703 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0408] The computer device may further include an input unit 704, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0409] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically in this embodiment, the processor 701 in the computer device will load the executable files corresponding to the processes of one or more application programs into the memory 702 according to the following instructions, and the processor 701 will run the application programs stored in the memory 702, thereby implementing the various method steps provided in the above embodiments, as follows:

[0410] Acquire the promotion position data of the promotion position, and perform binning processing according to the promotion position data to obtain the promotion position value data after binning processing; acquire the target object data of the target object, and perform binning processing according to the target object data to obtain the object quality data after binning processing; combine the promotion position value data and the object quality data to obtain target combination data; perform binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, acquire the adjustment weight of the target object under the promotion position, and determine the corresponding candidate queue length based on the adjustment weight.

[0411] In the above embodiments, the description of each embodiment has its own focus. For the part that is not described in detail in a certain embodiment, please refer to the detailed description of the queue processing method above, and will not be repeated here.

[0412] As can be seen from the above, the computer device of the embodiment of the present application can obtain the promotion position data of the promotion position, and perform binning processing according to the promotion position data to obtain the promotion position value data after binning processing; obtain the target object data of the target object, and perform binning processing according to the target object data to obtain the object quality data after binning processing; combine the promotion position value data and the object quality data to obtain the target combination data; perform binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, obtain the adjustment weight of the target object under the promotion position, and determine the corresponding candidate queue length based on the adjustment weight. In this way, the promotion position value data of the promotion position and the object quality data of the target object are combined to process the candidate queue length, so as to realize the allocation of computing resources taking into account the data characteristics of the promotion position and the data characteristics of the target object at the same time. Compared with the scheme in the related art that only sets computing resources according to the promotion position value of the promotion position, the embodiment of the present application greatly improves the accuracy of computing resource allocation, thereby improving the utilization rate of computing resources.

[0413] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0414] To this end, an embodiment of the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any queue processing method provided in the embodiment of the present application. For example, the instructions can execute the following steps:

[0415] Acquire the promotion position data of the promotion position, and perform binning processing according to the promotion position data to obtain the promotion position value data after binning processing; acquire the target object data of the target object, and perform binning processing according to the target object data to obtain the object quality data after binning processing; combine the promotion position value data and the object quality data to obtain target combination data; perform binning processing according to the target combination data to obtain the target value data after binning processing; based on the target binning interval where the target value data is located, acquire the adjustment weight of the target object under the promotion position, and determine the corresponding candidate queue length based on the adjustment weight.

[0416] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in various optional implementations provided in the above embodiments.

[0417] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0418] The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0419] Since the instructions stored in the computer-readable storage medium can execute the steps in any queue processing method provided in the embodiments of the present application, the beneficial effects that can be achieved by any queue processing method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0420] The above is a detailed introduction to a queue processing method, device, and computer-readable storage medium provided in an embodiment of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A queue processing method, It is characterized in that include: Acquire promotion position data of the promotion position, and perform binning processing according to the promotion position data to obtain promotion position value data after binning processing; Acquire target object data of the target object, and perform binning processing according to the target object data to obtain object quality data after binning processing; Combining the promotion position value data and the object quality data to obtain target combination data; Perform binning processing according to the target combination data to obtain target value data after binning processing; Based on the target binning interval in which the target value data is located, the adjustment weight of the target object under the promotion position is obtained, and the corresponding candidate queue length is determined based on the adjustment weight.

2. The queue processing method according to claim 1, It is characterized in that The step of acquiring promotion position data of the promotion position, and performing binning processing according to the promotion position data to obtain promotion position value data after binning processing includes: Obtain consumption data of promotional slots and the number of promotional slot requests; Calculate the cost per thousand requests data corresponding to the promotion position according to the consumption data and the promotion position request volume; The thousand-times-requests cost data is used as promotion position data, and equal-frequency binning is performed according to the promotion position data; The promotion position value data of the promotion position is determined according to the promotion position bin interval in which the promotion position data is located.

3. The queue processing method according to claim 2, It is characterized in that The equal-frequency binning process according to the promotion position data includes: Acquire first exposure data associated with the promotion position data, and perform statistics on the first exposure data to obtain first total exposure data; Determine a first exposure bin threshold point according to the first total exposure data and a first preset bin number; Based on the first exposure binning threshold point, the promotion position data is subjected to equal-frequency binning processing.

4. The queue processing method according to claim 3, It is characterized in that The performing equal-frequency binning processing on the promotion position data based on the first exposure binning threshold point includes: Sorting the promotion position data in ascending order; Accumulate the first exposure data associated with the sorted promotion position data one by one according to the sorting, to obtain the first accumulated exposure data corresponding to each sorted promotion position data; The promotion position data corresponding to the first accumulated exposure data that first reaches the first exposure bin threshold point is used as the promotion position bin interval threshold, and the promotion position bin interval is determined according to the promotion position bin interval threshold; The promotion position data is binned based on the promotion position binning intervals.

5. The queue processing method according to any one of claims 1 to 4, It is characterized in that The step of acquiring target object data of the target object and performing binning processing according to the target object data to obtain binned object quality data includes: Get the CPM data of the target object; The thousand-times-exposure cost data is used as target object data, and equal-frequency binning processing is performed according to the target object data; The object quality data of the target object is determined according to the target object binning interval in which the target object data is located.

6. The queue processing method according to claim 5, It is characterized in that The equal-frequency binning process according to the target object data includes: Acquire second exposure data associated with the target object data, and perform statistics on the second exposure data to obtain second total exposure data; Determine a second exposure bin threshold point according to the second total exposure data and a second preset bin number; Based on the second exposure binning threshold point, equal-frequency binning is performed on the target object data.

7. The queue processing method according to claim 6, It is characterized in that The performing equal-frequency binning processing on the target object data based on the second exposure binning threshold point includes: Sorting the target object data in ascending order; Accumulating the second exposure data associated with the sorted target object data one by one according to the sorting, to obtain second accumulated exposure data corresponding to each sorted target object data; Using the target object data corresponding to the second accumulated exposure data that first reaches the second exposure binning threshold point as the target object binning interval threshold, and determining the target object binning interval according to the target object binning interval threshold; The target object data is binned based on the target object binning interval.

8. The queue processing method according to claim 1, It is characterized in that The step of combining the promotion position value data and the object quality data to obtain target combination data includes: Combining the promotion position value data with the first weight to obtain target promotion position value data; Combining the object quality data with the second weight to obtain target object quality data; The target promotion position value data and the target object quality data are summed to obtain target combination data.

9. The queue processing method according to claim 1, It is characterized in that The method further comprises: Obtaining a target background load rate corresponding to the promotion position in a preset time period; The step of combining the promotion position value data and the object quality data to obtain target combination data includes: Summing the promotion position value data and the object quality data to obtain intermediate combination data; The ratio of the intermediate combination data to the target background load rate is calculated to obtain the target combination data.

10. The queue processing method according to claim 9, It is characterized in that The obtaining of the target background load rate corresponding to the promotion position in the preset time period includes: Obtaining a first background load rate of the promotion position in a preset time period and a second background load rate in a historical preset time period; Combining the first background load rate with the third weight to obtain a first target background load rate; Combining the second background load rate with the fourth weight to obtain a second target background load rate; The first target background load rate and the second target background load rate are summed to obtain the target background load rate of the promotion position.

11. The queue processing method according to claim 9, It is characterized in that The step of summing the promotion position value data and the object quality data to obtain intermediate combination data includes: Combining the promotion position value data with the fifth weight to obtain intermediate promotion position value data; combining the object quality data and the sixth weight to obtain intermediate object quality data; The intermediate promotion position value data and the intermediate object quality data are summed to obtain intermediate combination data.

12. The queue processing method according to claim 1, It is characterized in that The binning process is performed according to the target combination data to obtain the target value data after binning process, including: Perform binning processing according to the target combination data; Based on the target binning interval in which the target combination data is located, target value data of the target combination data is determined.

13. The queue processing method according to claim 1, It is characterized in that The determining the corresponding candidate queue length based on the adjustment weight includes: Obtaining the initial queue length under the promotion position; The product of the initial queue length and the adjustment weight is calculated to determine the candidate queue length of the target object under the promotion position.

14. A queue processing device, It is characterized in that include: A first binning processing unit is used to obtain promotion position data of a promotion position, and perform binning processing according to the promotion position data to obtain promotion position value data after binning processing; A second binning processing unit is used to obtain target object data of the target object, and perform binning processing according to the target object data to obtain object quality data after binning processing; A combining unit, used for combining the promotional position value data and the object quality data to obtain target combination data; A third binning processing unit, configured to perform binning processing according to the target combination data to obtain target value data after binning processing; A determination unit is used to obtain an adjustment weight of the target object under the promotion position based on the target binning interval in which the target value data is located, and determine a corresponding candidate queue length based on the adjustment weight.

15. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the steps in the queue processing method according to any one of claims 1 to 13.

16. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the queue processing method according to any one of claims 1 to 13 are implemented.

17. A computer program product comprising a computer program or instructions, It is characterized in that When the computer program or instruction is executed by a processor, the steps in the queue processing method described in any one of claims 1 to 13 are implemented.

Images