Method, apparatus, electronic device, and computer program product for processing push information

Abstract

Embodiments of the present application disclose a method, apparatus, electronic device, and computer program product for processing push information. The method includes: obtaining at least one piece of push information to be pushed; determining a reserved resource value corresponding to each piece of push information according to a resource calculation ratio and competition information corresponding to each piece of push information; the resource calculation ratio is determined according to a first target value of a first push metric; determining a resource value to be deducted corresponding to each piece of push information according to an initial resource value corresponding to each piece of push information and the reserved resource value; and deducting the numerical resources of each piece of push information according to the resource value to be deducted corresponding to each piece of push information. The above method, apparatus, electronic device, and computer program product for processing push information can more accurately deduct the numerical resources of push information and ensure that the first target value of the first push metric is achieved.

Description

Technical Field

[0001] The present application relates to the field of Internet technologies, and particularly to a method, an apparatus, an electronic device, and a computer program product for processing push information. Background Art

[0002] In recent years, with the rapid development of the Internet, information push has become one of the key services of many Internet platforms. By pushing various types of information to user terminals, users can be guided to pay attention to products or perform user behaviors related to products, thereby helping the customers of Internet platforms generate revenues. During the process of an Internet platform pushing information to a user terminal, the Internet platform will deduct the numerical resources of the push information. Therefore, how to accurately deduct the numerical resources of push information has always been one of the key issues of concern in the industry. Summary of the Invention

[0003] Embodiments of the present application disclose a method, an apparatus, an electronic device, and a computer program product for processing push information, which can more accurately deduct the numerical resources of push information and ensure that a first target value of a first push index is achieved.

[0004] Embodiments of the present application disclose a method for processing push information, including:

[0005] Obtaining at least one push information to be pushed;

[0006] Determining a reserved resource value corresponding to each push information according to a resource calculation ratio and competition information corresponding to each push information; the resource calculation ratio is determined according to a first target value of a first push index;

[0007] Determining a resource value to be deducted corresponding to each push information according to an initial resource value corresponding to each push information and the reserved resource value;

[0008] Deducting the numerical resources of each push information according to the resource value to be deducted corresponding to each push information.

[0009] Embodiments of the present application disclose a device for processing push information, the device including:

[0010] An information obtaining module, configured to obtain at least one push information to be pushed;

[0011] A first resource determining module, configured to determine a reserved resource value corresponding to each push information according to a resource calculation ratio and competition information corresponding to each push information; the resource calculation ratio is determined according to a first target value of a first push index;

[0012] The second resource determination module is configured to determine the resource values to be deducted corresponding to the respective push messages according to the initial resource values corresponding to the respective push messages and the reserved resource values;

[0013] The deduction module is configured to deduct the numerical resources of the respective push messages according to the resource values to be deducted corresponding to the respective push messages.

[0014] An embodiment of the present application discloses an electronic device, including a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor implements the method described in any one of the above embodiments.

[0015] An embodiment of the present application discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0016] An embodiment of the present application discloses a computer program product, including a computer program. When the computer program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0017] The method, device, electronic device, storage medium, and computer program product for processing push messages disclosed in the embodiments of the present application obtain at least one push message to be pushed, determine the reserved resource values corresponding to the respective push messages according to the resource calculation ratio and the competition information corresponding to the respective push messages, determine the resource values to be deducted corresponding to the respective push messages according to the initial resource values and the reserved resource values corresponding to the respective push messages, and deduct the numerical resources of the respective push messages according to the resource values to be deducted corresponding to the respective push messages. By combining the initial resource values and the reserved resource values of the push messages to jointly determine the resource values to be deducted corresponding to the push messages, the numerical resources of the push messages can be deducted more accurately, and the reserved resource values are determined according to the resource calculation ratio and the competition information corresponding to the push messages. The resource calculation ratio is determined according to the first target value of the first push index, so as to ensure that the pushing of the push messages and the deduction of the numerical resources can achieve the first target value of the first push index. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1A It is an application scenario diagram of the method for processing push messages in an embodiment;

[0020] Figure 1B It is a system architecture diagram of a method for processing push information in an embodiment;

[0021] Figure 1C It is a schematic diagram of a processing flow of push information in an embodiment;

[0022] Figure 2 It is a flowchart of a method for processing push information in an embodiment;

[0023] Figure 3 It is a flowchart of determining the reserved resource value corresponding to each push information according to the resource calculation ratio and the competition information corresponding to each push information in an embodiment;

[0024] Figure 4 It is a flowchart of determining the estimated expected revenue value corresponding to the first ratio value in an embodiment;

[0025] Figure 5 It is a flowchart of a method for processing push information in another embodiment;

[0026] Figure 6 In an embodiment, sort multiple candidate push messages according to the competition information respectively corresponding to the multiple candidate push messages;

[0027] Figure 7 It is a flowchart of a method for processing push information in yet another embodiment;

[0028] Figure 8 It is a block diagram of a device for processing push information in an embodiment;

[0029] Figure 9 It is a block diagram of a device for processing push information in another embodiment;

[0030] Figure 10 It is a block diagram of an electronic device in an embodiment. Detailed implementation manners

[0031] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0032] It should be noted that the terms "include" and "have" and any variations thereof in the embodiments of the present application and the accompanying drawings are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0033] It can be understood that the terms "first", "second", etc. used in the present application may be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the present application, the first push metric may be referred to as the second push metric, and similarly, the second push metric may be referred to as the first push metric. Both the first push metric and the second push metric are push metrics, but they are not the same push metric. The term "plurality" used in the present application refers to two or more.

[0034] The following is an explanation and description of some terms that the present application may involve:

[0035] Push information: refers to the information pushed by the push platform to the client / user terminal, and the push information may include information flows such as advertisements.

[0036] Push location: refers to the location where the push information is displayed on the client, and the push location can be a spatial location or a temporal location. The spatial location can be the page location where the push information is displayed, etc., and the temporal location refers to the time when the push information is displayed.

[0037] Competition resource value: refers to the numerical resources that the customer corresponding to the push information is willing to pay in the competition for the push location, such as the bid of the advertiser in each competition for the advertising position.

[0038] Numerical resources: refer to resources represented by numerical values, such as money, price, etc.

[0039] CPM (Cost Per Mille, cost per thousand impressions): A charging method based on the number of impressions. For every 1000 impressions of the push information (such as an advertisement), a charge is made once.

[0040] CPC (Cost Per Click, cost per click): A charging method based on the number of clicks. For every time the push information is clicked by the user, a charge is made once.

[0041] CPD (Cost Per Download, cost per download): A charging method based on the number of downloads. For every time the product related to the push information (such as an application program, etc.) is downloaded, a charge is made once.

[0042] CPA (Cost Per Action): A billing method based on the actual delivery effect of the pushed information. The action cost may include, but is not limited to, download cost, installation cost, registration cost, questionnaire filling cost, etc. Different actions can be set with different fees.

[0043] oCPM (Optimized Cost per Mille): A thousand - impression charging method optimized for target conversion.

[0044] GMV (Gross Merchandise Volume): The sum of the order amounts concluded within a certain period of time, which can include both the paid and unpaid parts by users.

[0045] pCTR (Predict Click - Through Rate): The predicted probability that users will click on the pushed information.

[0046] pCVR (Predicted Conversion Rate): The probability of conversion after the pushed information is clicked.

[0047] Predicted exposure rate: The predicted probability that the pushed information will be displayed.

[0048] Predicted resource conversion volume: The predicted amount of conversion into numerical resources after the pushed information is displayed, such as the predicted transaction amount of the placed advertisement, etc.

[0049] ROI (Return Over Invest): The value returned through investment activities, such as the return that a customer can obtain by pushing information to users through the push platform.

[0050] First - price billing rule: In the competition for push positions, a sealed - bid auction is used for competition. The first - price billing rule means billing according to the customer's bid or according to the revenue of the pushed information itself.

[0051] Second - price billing rule: In the competition for push positions, a sealed - bid auction is used for competition. The second - price billing rule means billing according to the bid of the next pushed information in the ranking or the revenue of the next pushed information.

[0052] Linear programming problem: Also known as linear programming, in mathematics, linear programming can refer to an optimization problem where both the objective function and the constraints are linear.

[0053] Dual problem: Every linear programming problem has a dual problem. A linear programming problem (i.e., the primal problem) and its dual problem are described from different perspectives for an actual problem, forming a pair of mutually dual linear programming problems.

[0054] Dual variable: It refers to the variable in the dual problem. The constraint conditions of a linear programming problem correspond one-to-one with the dual variables, that is, the constraint conditions of one problem correspond to the variables of the other problem, and the variables of one problem correspond to the constraint conditions of the other problem.

[0055] KKT (Karush-Kuhn-Tucker Conditions) conditions: The conditions that need to be satisfied to obtain the optimal solution of an optimization problem under given constraint conditions.

[0056] Complementary slackness condition: A condition in the KKT conditions, derived from the strong duality between the primal problem and the dual problem (which means that the optimal solutions of the primal problem and the dual problem can both be achieved and are equal), and can be used to obtain the optimal solution of the primal problem and / or the dual problem.

[0057] Hyperparameter: It refers to the parameters preset in the model. Usually, hyperparameters need to be optimized to select a set of optimal hyperparameters to improve the performance of the model.

[0058] Expected platform revenue: The estimated revenue of the push platform for information push. The expected platform revenue can be associated with the customer's bid, the estimated click-through rate of the pushed information, the estimated exposure rate, etc.

[0059] In recent years, with the rapid development of the Internet, information push has become one of the key businesses of many Internet platforms. Taking the online advertising provided by Internet platforms as an example, advertisers can attract network audiences to pay attention to relevant products and achieve product promotion by placing online ads. According to different billing methods, online ads can be divided into different types such as CPM, CPC, CPD, oCPM, etc. For example, if a customer places a CPC ad, the ad will be billed at the click link, and the ad delivery platform will charge the customer once for each click of the ad; for CPM ads, the ad delivery platform will charge according to the number of ad exposures.

[0060] In addition, for e-commerce ads and non-e-commerce ads, the key concerns of customers are different. For e-commerce ads, customers are more concerned about ROI, while the ad delivery platform needs to pay attention to various indicators such as user experience, platform revenue, and profit in addition to ROI; for non-e-commerce ads, customers are more concerned about whether the actual conversion cost exceeds the expected conversion cost and the amount of conversions, while the ad delivery platform focuses on whether the actual conversion cost is lower than the expected conversion cost and the actual deducted amount, etc.

[0061] Most online advertisements adopt the method of real-time bidding. Each time an advertisement is placed, the advertisement placement platform will select an advertisement that matches the user for placement. If there are multiple matching advertisements, real-time bidding will be carried out. The advertisement placement platform can calculate the competitiveness of each advertisement and perform a bidding ranking, so as to place advertisements according to the result of the bidding ranking. Since the concerns of e-commerce advertisements and non-e-commerce advertisements are different, generally, the competitiveness of e-commerce advertisements can be calculated using Equation (1):

[0062] ecpm = α * pctr + β * pctr * pcvr * pAmount Equation (1);

[0063] Among them, ecpm represents the revenue per thousand impressions, which refers to the revenue that an advertisement can obtain for every thousand impressions. Ecpm can be used as the competitiveness of the advertisement; pctr is the predicted click-through rate of the advertisement; pcvr is the predicted conversion rate of the advertisement; pAmount is the predicted transaction amount of the advertisement. Among them, the first item in Equation (1) considers the user experience, and the second item considers the customer's GMV. The α parameter and the β parameter can be set according to experience.

[0064] For non-e-commerce advertisements, the competitiveness of the advertisement can be calculated using Equation (2):

[0065] ecpm = pctr * pcvr * CPA Equation (2);

[0066] Among them, CPA represents the bid price of the customer for each behavior of the user.

[0067] In the related technology, the calculation method of the competitiveness of advertisements is difficult to take into account the needs of multiple parties such as customers, user experience, and advertisement placement platforms at the same time, and it is impossible to consider the competitiveness of advertisements from a global perspective, resulting in inaccurate evaluation of the competitiveness of advertisements and poor advertisement placement effects.

[0068] For e-commerce advertisements and non-e-commerce advertisements, the billing method of the advertisement placement platform is usually the same. For example, the second-price billing rule can be used for billing. In the same advertisement bidding, according to the order of the competitiveness of the advertisements from large to small, multiple advertisements can be ranked. Among them, the billing of the advertisement can be shown as Equation (3):

[0069] price n = ecpm n+1 / pctr n Equation (3);

[0070] Among them, price n represents the billing of the nth advertisement, ecpm n+1 represents the ecpm of the advertisement ranked next to the nth advertisement, pctrn Represents the estimated click-through rate of the nth advertisement.

[0071] In the second-price billing rule, each advertisement billing is determined by the eCPM of the next advertisement after sorting and its own estimated click-through rate. The eCPM of the next advertisement and its own eCPM may vary greatly or slightly, so it will lead to inaccurate billing and it is difficult to ensure the advertising delivery effect or cost deviation. For example, the situation where the ROI is difficult to achieve may occur.

[0072] The embodiments of the present application disclose a method, device, electronic device, and storage medium for processing push information, which can more accurately deduct the numerical resources of push information and ensure that the first target value of the first push index is achieved.

[0073] Figure 1A It is an application scenario diagram of the method for processing push information in an embodiment. As Figure 1A shown, the method for processing push information can be applied to the push platform 110, which can be a server or a server cluster. The push platform 110 can be communicatively connected to multiple user terminals 120.

[0074] Among them, the user terminal 120 refers to the terminal used by the user. The user terminal 120 can include but is not limited to mobile phones, wearable devices (such as smart glasses, smart watches, etc.), tablet computers, vehicle-mounted terminals, laptop computers, PCs (Personal Computers, personal computers), etc. The push platform 110 can send push information to the user terminal 120, and the user terminal 120 can display the push information.

[0075] In some embodiments, the push platform 110 can also be communicatively connected to multiple customer terminals 130.

[0076] The customer terminal 130 can refer to the terminal used by the customer. The customer refers to the merchant who publishes the push information, such as the advertiser who publishes the advertisement. The customer terminal 130 can include but is not limited to mobile phones, wearable devices, tablet computers, vehicle-mounted terminals, laptop computers, PCs, etc. The customer can set the push strategy, competition information, view the push effect, etc. of the push information through the customer terminal 130. The customer terminal 130 can perform data transmission with the push platform 110, and can upload the set push strategy, competition information, etc. of the push information to the push platform 110, and can also obtain the push effect of the push information from the push platform 110, such as click-through rate, conversion rate, etc.

[0077] It should be noted that the present application embodiment does not limit the establishment method of the communication connection between the push platform 110 and the user terminal 120, the customer terminal 130, etc. For example, communication can be carried out based on HTTP (Hyper Text Transformer Protocol, Hypertext Transfer Protocol), HTTPS (Hypertext Transfer Protocol Secure, Hypertext Transfer Security Protocol), WebSocket protocol, etc.

[0078] Figure 1B It is a system architecture diagram of the push information processing method in an embodiment. As Figure 1B shown, a database 112 and an algorithm engine 114 can be deployed on the push platform 110. The database can be used to store a large amount of push information and data related to the push information. The algorithm engine can be used to implement real-time competition calculation of push information, numerical resource deduction calculation, etc.

[0079] When the user accesses the client, such as a website or an APP (Application, application program), through the user terminal 120, the client can send a competition request to the push platform 110. The competition request can include user information, push location, etc. The algorithm engine 114 of the push platform 110 can select multiple candidate push information according to the competition request, sort them according to the competition information corresponding to the multiple candidate push information, and then determine the push information to be pushed corresponding to each push location based on the sorted multiple candidate push information.

[0080] After the algorithm engine 114 of the push platform 110 determines the push information to be pushed corresponding to each push location, it can also calculate the resource value to be deducted corresponding to each push information to be pushed according to the resource calculation ratio and the competition information corresponding to each push information to be pushed, and then deduct the numerical resources of each push information according to the resource value to be deducted corresponding to each push information to be pushed. The resource calculation ratio is determined according to the first target value of the first push index, so as to ensure that the information push of the push platform 110 can achieve the first target value of the first push index.

[0081] The push platform 110 can feedback a competition response to the client. The competition response can include information such as the identifiers of each push information to be pushed. The client can obtain information such as the identifiers of each push information to be pushed according to the competition response, and can access the database of the push platform 110 based on the identifiers of each push information to be pushed to obtain the corresponding push information and display it.

[0082] Exemplarily, Figure 1C It is a schematic diagram of the push information processing flow in an embodiment. As Figure 1CAs shown, after the push platform 110 obtains a competition request (such as an advertising bidding request), it performs push message recall. Push message recall can be understood as selecting multiple push messages that match the current traffic from all push messages. For example, if the competition request is generated after the user enters a search keyword on the client, multiple push messages that match the search keyword can be selected from all push messages; or, for another example, multiple push messages that match the user information (such as user habits, age, gender, etc.) can also be selected from all push messages. After the push message recall, a rough ranking can be performed on the multiple push messages obtained in the recall stage, roughly sorting and filtering the multiple push messages obtained in the recall stage to obtain a smaller number of candidate push messages, and then performing a fine ranking on the smaller number of candidate push messages, accurately calculating the competition scores corresponding to each candidate push message and performing ranking, and selecting the push messages to be pushed corresponding to each push position. The sorting mechanism involved in the embodiments of the present application can be applied to this fine ranking stage. The amount of processed data in the recall stage, rough ranking stage, and fine ranking stage gradually decreases, and the calculation accuracy gradually increases, so as to ensure the accuracy of the push and take into account the operation pressure.

[0083] After the fine ranking stage, the push platform 110 can perform numerical resource deduction on each push message to be pushed. In the embodiments of the present application, the numerical resource deduction can be performed by combining the initial resource value and the reserved resource value of the push message. In some embodiments, the push platform 110 can also perform mixed ranking, which means mixing and ranking the push messages to be pushed determined in the fine ranking stage with other types of push messages. For example, advertisements can be mixed with other types of push messages, and finally the mixed push results are displayed on the client, thereby improving the user experience.

[0084] It should be noted that Figure 1C only one processing flow of push messages is shown. In actual applications, the processing flow of push messages may include Figure 1C more or fewer processing stages. For example, the rough ranking stage or the recall stage can be omitted, or a rearrangement stage can be added, etc. The embodiments of the present application do not limit this.

[0085] For example Figure 2 As shown, in one embodiment, a method for processing push messages is provided, which can be applied to the above-mentioned push platform. The method may include the following steps:

[0086] Step 210, obtain at least one push message to be pushed.

[0087] The push platform can obtain at least one push message to be pushed. The push message to be pushed can refer to the push message prepared to be pushed to the client for display, that is, the push message that has been determined to be pushed. Taking the push message as an advertisement as an example, the push message to be pushed can be the advertisement to be placed. Each time the push platform receives a competition request sent by the client, it can select the push messages to be pushed corresponding to each push position included in the competition request according to the competitiveness of each push message, and push them to the client. Optionally, the push position and the push message to be pushed can be in a one-to-one correspondence relationship, and the push platform can obtain the push messages to be pushed corresponding to each push position in the current competition request.

[0088] As an implementation manner, when the push platform obtains at least one push message to be pushed, it can be to obtain the identifiers corresponding to each push message to be pushed. The identifier can be used to identify the push message, and the identifier can be composed of one or more of numbers, letters, symbols, etc. For example, the identifier can be the number of the push message, etc., but is not limited thereto.

[0089] Furthermore, the competition information corresponding to each push message to be pushed can also be obtained. The competition information can be used to describe the competitiveness of the push message. The competition information can include the competition resource value and the estimated display information, etc. The estimated display information can include one or more of the estimated click-through rate, the estimated conversion rate, and the estimated resource conversion amount, etc.

[0090] Optionally, the competition information corresponding to the push message can be configured by the customer corresponding to the push message on the customer terminal. The customer terminal can adopt the method of configuring the competition information in real time, or can also adopt the method of configuring the competition information at intervals. The method of configuring the competition information in real time can refer to that the customer terminal configures the competition information in each competition request. For example, when the push message receives a competition request each time, it can send a competition notice to the customer terminals corresponding to each push message. After receiving the competition notice, the customer terminal can configure the competition information for this time, such as configuring the competition resource value, etc., and then upload the configured competition information to the push platform. The method of configuring the competition information at intervals can refer to that the customer terminal configures the competition information at intervals. The interval time period can be a fixed time period or an unfixed time period. After the customer terminal reconfigures the competition information each time, it can upload the configured competition information to the push platform, and the push platform can store the latest competition information of the push message. Optionally, the competition information corresponding to the push message can also be generated by the push platform according to the characteristics of the push message. For example, the push platform can generate the estimated click-through rate, the estimated conversion rate, etc. corresponding to the push message according to the characteristics of the push message.

[0091] Step 220: Determine the reserved resource value corresponding to each push message according to the resource calculation ratio and the competition information corresponding to each push message.

[0092] The resource value to be deducted corresponding to each push message to be pushed can be determined according to the competition information and resource calculation ratio corresponding to each push message to be pushed. In the embodiments of the present application, in the process of determining the resource value to be deducted corresponding to each push message to be pushed, the initial resource value of the push message to be pushed can be combined with the reserved resource value. The initial resource value may refer to the initial resource value determined by using the traditional calculation method of the resource value to be deducted. For example, the initial resource value may be the resource value to be deducted calculated by using the second-price billing rule, etc., but is not limited thereto. The reserved resource value can be used to represent the minimum value of the push message. In the embodiments of the present application, the reserved resource value may be dynamically changed, and the reserved resource values corresponding to different push messages may be different, and the reserved resource values determined for the same push message in each competition request may also be different.

[0093] The reserved resource value corresponding to each push message can be determined according to the resource calculation ratio and the competition information corresponding to each push message. The resource calculation ratio can be determined according to the first target value of the first push metric. The first push metric can be a metric for measuring the overall push effect of the push platform. The first target value can be preset, which is the target that the push platform expects to achieve in the first push metric. Exemplarily, the first push metric may include the overall ROI of the push platform, then the first target value corresponding to the first push metric can be the preset target return on investment, that is, the target ROI. It should be noted that the first push metric can also be other metrics, such as the push cost deviation corresponding to the push platform, etc., which is not limited herein.

[0094] The resource calculation ratio can be adjusted according to the first target value of the first push metric. Optionally, the resource calculation ratio and the first target value of the first push metric may be in a negative correlation relationship. The larger the first target value of the first push metric, the smaller the resource calculation ratio; the smaller the first target value of the first push metric, the larger the resource calculation ratio. Optionally, the resource calculation ratio and the first target value of the first push metric may also be in a positive correlation relationship. The larger the first target value of the first push metric, the larger the resource calculation ratio; the smaller the first target value of the first push metric, the smaller the resource calculation ratio, which is not limited herein.

[0095] In some embodiments, the resource calculation ratio can be adjusted according to a fixed time period. The fixed time period can be set according to actual needs. For example, the resource calculation ratio can be adjusted once every day, or can be adjusted once every 2 days, 5 days, 1 week, etc. time, which is not limited herein.

[0096] In some embodiments, the resource calculation ratio may be determined according to the first target value of the first push metric and historical push data. The historical push data may include the total resource conversion amount corresponding to the historical time period, the actual display information corresponding to multiple historical push messages pushed during the historical time period, and the deducted numerical resources.

[0097] Among them, the historical time period can be set according to actual needs, such as the past 1 day, the past 2 days, the past 5 days, the past 1 week, etc., but not limited thereto. Further, the historical time period may be the same as the above-mentioned fixed time period. For example, if the resource calculation ratio is adjusted every 1 day, the historical push data for the past 1 day can be obtained to determine the resource calculation ratio.

[0098] The total resource conversion amount corresponding to the historical time period may refer to the sum of the resource conversion amounts corresponding to the push messages pushed by the push platform during the historical time period. For example, the total resource conversion amount corresponding to the historical time period may be the GMV corresponding to the historical time period.

[0099] The multiple historical push messages pushed during the historical time period may include all the historical push messages pushed in each competition request during the historical time period. Each historical push message can be marked with the corresponding competition request serial number and the arrangement serial number in the competition request. For example, the b-th push message in the a-th competition request during the historical time period can be marked with "ab". It should be noted that in different competition requests, there may be the same historical push message, but the markings of this historical push message in different competition requests are different. The historical push data will include data such as the estimated display information and deducted numerical resources of this historical push message in each competition request.

[0100] The actual display information corresponding to the historical push message may refer to the data actually generated after the historical push message is pushed. The actual display information may include one or more of, but not limited to, the actual click-through rate, the actual exposure rate, the actual conversion rate, etc.

[0101] The deducted numerical resources corresponding to the historical push message may refer to the deducted numerical resources of the historical push message.

[0102] According to the historical push data and the first target value of the first push metric, the resource calculation ratio corresponding to achieving the first target value of the first push metric can be calculated, and then the reserved resource value of the push message to be pushed can be calculated using this resource calculation ratio, so as to ensure that the push of the push message to be pushed and the deduction of numerical resources can achieve the first target value of the first push metric.

[0103] In some embodiments, the first target value of the first push metric includes the target ROI, and the estimated display information corresponding to the historical push information may include the actual click-through rate corresponding to the historical push information. The resource calculation ratio may be determined according to the first ratio and the actual click-through rates and deducted numerical resources corresponding to multiple historical push information respectively, and the first ratio may be the ratio between the total resource conversion amount corresponding to the historical time period and the target return on investment.

[0104] In the historical time period, the actual click-through rate, deducted numerical resource, and the resource calculation ratio to be calculated of each historical push information for each competition request can be multiplied, and then all the product results are accumulated. The accumulated result can be made equal to the first ratio, so as to calculate the resource calculation ratio. As a specific implementation manner, the resource calculation ratio can be calculated using Equation (4):

[0105] ∑priceRate*ctr ij *price ij =GMV / R1 Equation (4);

[0106] where priceRate represents the resource calculation ratio, ctr ij represents the actual click-through rate of the j-th historical push information in the i-th competition request in the historical time period, price ij represents the deducted numerical resource of the j-th historical push information in the i-th competition request in the historical time period, GMV represents the total resource conversion amount corresponding to the historical time period, and R1 represents the target ROI. The push platform can adjust the target ROI according to actual needs, thereby adjusting the resource calculation ratio, and jointly determining the resource calculation ratio with the historical push data of the result historical time period, which can improve the accuracy of the calculated resource calculation ratio, more accurately ensure that the push of the push information to be pushed and the deduction of numerical resources can achieve the target ROI, and at the same time protect the interests of the push platform and customers.

[0107] It should be noted that other actual display information of the historical push information can also be collected to determine the resource calculation ratio, which can be adjusted according to business requirements. For example, in the CPC business model, the actual click-through rate of the historical push information can be used to determine the resource calculation ratio, and in the CPM business model, the actual exposure rate of the historical push information can be used to determine the resource calculation ratio, etc., but not limited thereto.

[0108] As an implementation manner, the reserved resource value corresponding to each push message to be pushed can be determined according to the resource calculation ratio and the competition resource value corresponding to each push message to be pushed. For example, the resource calculation ratio can be multiplied by the competition resource value corresponding to each push message to be pushed to obtain the reserved resource value corresponding to each push message to be pushed; alternatively, the reserved ratio corresponding to the current competition request can be determined according to the resource calculation ratio, and then the reserved ratio can be multiplied by the competition resource value corresponding to each push message to be pushed to obtain the reserved resource value corresponding to each push message to be pushed. The reserved ratio can refer to the ratio of reserving the competition resource value corresponding to the push message. The reserved resource value calculated using the resource calculation ratio is more accurate, which can improve the accuracy of deducting the numerical resources of the push message.

[0109] Step 230: Determine the resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message.

[0110] Step 240: Deduct the numerical resources of each push message according to the resource value to be deducted corresponding to each push message.

[0111] After determining the reserved resource value corresponding to each push message to be pushed, the reserved resource value corresponding to each push message to be pushed and the initial resource value can be combined to jointly determine the resource value to be deducted corresponding to each push message to be pushed. As an implementation manner, the initial resource value corresponding to each push message to be pushed can be the initial resource value calculated using the second-price billing method. In the case of multiple push messages to be pushed, the multiple push messages can be arranged in order, and the initial resource value corresponding to each push message to be pushed can be related to the next push message. Taking the second push message as an example, the second push message can be at least one of the multiple push messages to be pushed. The initial resource value corresponding to the second push message can be determined according to the estimated display information of the second push message and the display revenue corresponding to the next push message of the second push message.

[0112] The display revenue corresponding to the next push message can include, but is not limited to, the revenue per thousand impressions of the next push message. Further, the initial resource value corresponding to the second push message can be equal to the revenue per thousand impressions of the next push message divided by the estimated click-through rate of the second push message. Specifically, the revenue per thousand impressions corresponding to the next push message of the second push message can be calculated first according to formula (1) or formula (2) above, and then the initial resource value corresponding to the second push message can be calculated through formula (3). Calculating the initial resource value corresponding to the push message to be pushed using the second-price billing method can create a better push competition environment and ensure the quality of the push message and the interests of the push platform.

[0113] As an implementation manner, the maximum value between the reserved resource value and the initial resource value corresponding to the push message to be pushed may be taken as the resource value to be deducted corresponding to the push message to be pushed. Specifically, the resource value to be deducted corresponding to the push message to be pushed may be determined by using Equation (5):

[0114] price j =max(gsp j ,P j ) Equation (5);

[0115] wherein, gsp j represents the initial resource value corresponding to the j-th push message to be pushed, P j represents the reserved resource value corresponding to the j-th push message to be pushed, and price j represents the resource value to be deducted corresponding to the j-th push message to be pushed.

[0116] After determining the resource value to be deducted corresponding to the push message to be pushed, the numerical resources of the push message to be pushed may be deducted according to the resource value to be deducted. The numerical resources deducted for the push message to be pushed will be adjusted in each competition request. By combining the initial resource value and the reserved resource value of the push message to be pushed, the problem of inaccurate deduction caused by only relying on the initial resource value for resource deduction can be reduced, and the risk of the push platform can be lowered.

[0117] In the embodiments of the present application, at least one push message to be pushed is obtained, the reserved resource value corresponding to each push message is determined according to the resource calculation ratio and the competition information corresponding to each push message, the resource value to be deducted corresponding to each push message is determined according to the initial resource value and the reserved resource value corresponding to each push message, and the numerical resources of each push message are deducted according to the resource value to be deducted corresponding to each push message. By combining the initial resource value and the reserved resource value of the push message, the resource value to be deducted corresponding to the push message is jointly determined, so that the numerical resources of the push message can be deducted more accurately. Moreover, the reserved resource value is determined according to the resource calculation ratio and the competition information corresponding to the push message, and the resource calculation ratio is determined according to the first target value of the first push index, so as to ensure that the push of the push message and the deduction of the numerical resources can achieve the first target value of the first push index.

[0118] As Figure 3 shown, in one embodiment, the step of determining the reserved resource value corresponding to each push message according to the resource calculation ratio and the competition information corresponding to each push message may include the following steps:

[0119] Step 302, determining the target expected revenue according to the resource calculation ratio and the competition information corresponding to at least one push message respectively.

[0120] The push platform can obtain the push list corresponding to the current competition request. The push list can include at least one push position and the push information to be pushed corresponding to each push position. When the push list includes multiple pieces of push information to be pushed, the multiple pieces of push information to be pushed can be arranged in order. The target expected revenue corresponding to the current competition request can be determined according to the resource calculation ratio and the competition information corresponding to each piece of push information included in the push list. The target expected revenue can refer to the revenue estimated to be generated by the push platform for pushing information for the current competition request.

[0121] In some embodiments, the first expected revenue corresponding to each piece of push information can be determined according to the resource calculation ratio, the competition resource value corresponding to each piece of push information, and the estimated display information, and then the target expected revenue can be determined according to the first expected revenue corresponding to each piece of push information.

[0122] The push platform can first determine the first expected revenue corresponding to each piece of push information to be pushed according to the resource calculation ratio, the competition resource value corresponding to each piece of push information to be pushed, and the estimated display information. The first expected revenue corresponding to the push information to be pushed can refer to the revenue estimated to be generated by the push platform for pushing the push information to be pushed in the current competition request.

[0123] The estimated display information can include, but is not limited to, one or more of estimated click-through rate, estimated conversion rate, estimated exposure rate, etc.

[0124] Optionally, the resource calculation ratio, the competition resource value corresponding to each piece of push information to be pushed, and the estimated click-through rate can be multiplied to obtain the first expected revenue corresponding to each piece of push information to be pushed, and then the first expected revenues corresponding to each piece of push information to be pushed can be accumulated to obtain the target expected revenue. Specifically, the target expected revenue can be determined by using formula (6):

[0125] S = sum(ctr j * bid j * priceRate) Formula (6);

[0126] where S represents the target expected revenue, ctr j represents the estimated click-through rate corresponding to the j-th piece of push information to be pushed in the push list, bid j represents the competition resource value corresponding to the j-th piece of push information to be pushed in the push list, and priceRate represents the resource calculation ratio. By using this method, the accuracy of the determined target expected revenue can be improved, thereby further improving the accuracy of the subsequent calculation of the retention ratio and the accuracy of the determined retained resource value.

[0127] Step 304: Determine the retention ratio according to the target expected revenue.

[0128] After determining the target expected revenue corresponding to the current competition request, the retention ratio corresponding to the current competition request can be determined according to the target expected revenue, so that the estimated expected revenue generated by the retention ratio can be the same as or basically match the target expected revenue.

[0129] In some embodiments, within the ratio range, a ratio value whose estimated expected revenue matches the target expected revenue can be searched for, and the matching ratio value can be used as the retention ratio. Among them, the estimated expected revenue corresponding to the first ratio value is determined according to the first ratio value and the competition information corresponding to at least one push message, and the first ratio value is any ratio value within the ratio range.

[0130] The ratio values included in the ratio range can be searched for, and the estimated expected revenue corresponding to each ratio value can be determined one by one until a ratio value whose estimated expected revenue matches the target expected revenue is found. The ratio range can be set in advance according to actual needs, such as 1% - 100%, etc., but is not limited thereto.

[0131] Exemplarily, the dichotomy method can be used to search for a ratio value within the ratio range whose estimated expected revenue matches the target expected revenue. The process of using the dichotomy method for searching can be to use the ratio range as the initial target search range, and the ratio value located in the middle of the target search range can be taken. According to this middle ratio value and the competition information corresponding to each push message to be pushed in the push list, the estimated expected revenue value corresponding to this middle ratio value can be determined, and the estimated expected revenue value can be compared with the target expected revenue to determine whether they match. If they do not match, and the estimated expected revenue value is greater than the target expected revenue, then the part of the target search range smaller than this middle ratio value can be used as the new target search range; if they do not match and the estimated expected revenue value is greater than the target expected revenue, then the part of the target search range larger than this middle ratio value can be used as the new target search range, and then continue to execute the step of taking the ratio value located in the middle of the target search range and determining the estimated expected revenue value corresponding to this middle ratio value until the estimated expected revenue value of this middle ratio value matches the target expected revenue.

[0132] Searching in the way of the dichotomy method can improve the search efficiency and reduce the computational complexity. It should be noted that other search methods can also be used for searching, such as sequential search, hash search, etc., which are not limited herein.

[0133] In some embodiments, as Figure 4 shown, taking the first ratio value in the ratio range as an example, determining the estimated expected revenue value corresponding to the first ratio value may include steps 402 - 406.

[0134] Step 402: Determine the first resource value corresponding to the first push message according to the first ratio value and the competing resource value corresponding to the first push message, where the first push message is any push message.

[0135] For each push message to be pushed included in the push list, the first resource value corresponding to each push message to be pushed can be calculated respectively according to the first ratio value and the competing resource value corresponding to each push message to be pushed. Taking the first push message in the push list as an example, the first ratio value can be multiplied by the competing resource value corresponding to the first push message to obtain the first resource value corresponding to the first push message. This first resource value can be understood as the reserved resource value calculated by taking the first ratio value as the reservation ratio.

[0136] Step 404: Determine the second resource value corresponding to the first push message according to the first resource value corresponding to the first push message and the initial resource value.

[0137] In some embodiments, the maximum value between the first resource value corresponding to the first push message and the initial resource value can be taken as the second resource value corresponding to the first push message. This second resource value can be understood as the resource value to be deducted corresponding to the first push message determined by taking the first ratio value as the reservation ratio. As a specific implementation manner, the second resource value corresponding to the push message to be pushed can be calculated by formula (7):

[0138] p' j =max(gsp j ,bid j *l) Formula (7);

[0139] where p' j represents the second resource value corresponding to the j-th push message to be pushed in the push list, bid j *l represents the first resource value corresponding to the j-th push message to be pushed in the push list, gsp j represents the initial resource value corresponding to the j-th push message to be pushed in the push list, and l is a ratio value (such as the first ratio value above).

[0140] Step 406: Determine the estimated expected revenue corresponding to the first ratio value according to the second resource values corresponding to at least one push message respectively and the corresponding estimated display information.

[0141] After determining the second resource value corresponding to each push message to be pushed, the estimated expected revenue corresponding to the first ratio value can be determined according to the second resource value corresponding to each push message to be pushed and the corresponding estimated display information. As an implementation, the second resource value corresponding to each push message to be pushed can be multiplied by the estimated click-through rate, and the multiplication results can be accumulated to obtain the estimated expected revenue corresponding to the first ratio value. Specifically, the estimated expected revenue corresponding to the first ratio value can be calculated using Equation (8):

[0142] s = sum(ctr j * bid j * p' j ) Equation (8);

[0143] where s represents the estimated expected revenue, ctr j represents the estimated click-through rate corresponding to the j-th push message to be pushed in the push list, bid j represents the competitive resource value corresponding to the j-th push message to be pushed in the push list, and p' j represents the second resource value corresponding to the j-th push message to be pushed in the push list. The estimated expected revenue s can be compared with the above-mentioned target expected revenue S to determine whether they match.

[0144] Optionally, the estimated expected revenue value matching the target expected revenue can be that the estimated expected revenue value is equal to the target expected revenue, or the absolute value of the difference between the estimated expected revenue value and the target expected revenue is less than the difference threshold, etc.

[0145] In the embodiments of the present application, the estimated expected revenue corresponding to the ratio value can be calculated according to each push message to be pushed, the retention ratio can be accurately searched, and the accuracy of calculating the retention resource value of each push message to be pushed subsequently can be further improved. Moreover, the retention ratio determined based on the target expected revenue can ensure that the pushing of the push message and the deduction of the numerical resources can achieve the first target value of the first push index (such as the target ROI, etc.), while taking into account the interests of both the push platform and the customer.

[0146] Step 306: Determine the retention resource value corresponding to each push message according to the retention ratio and the competitive information corresponding to each push message.

[0147] After determining the retention ratio corresponding to the current competition request, the retention ratio can be multiplied by the competitive resource value corresponding to each push message to be pushed included in the push list to obtain the retention resource value corresponding to each push message to be pushed. Specifically, the retention resource value can be calculated using Equation (9):

[0148] P j = bid j*lambda expression (9);

[0149] Where P j represents the reserved resource value corresponding to the j-th push message to be pushed in the push list, bid j represents the competing resource value corresponding to the j-th push message to be pushed in the push list, and lambda represents the reservation ratio.

[0150] In the embodiments of the present application, for each competition request, the reserved resource value of each push message to be pushed can be determined according to the resource calculation ratio and the competition information of each push message to be pushed corresponding to each competition request, and data resource adjustment for deducting the push message at the request level is performed, improving the accuracy of deducting the numerical resources of the push message, and ensuring that the pushing of the push message and the deduction of the numerical resources can achieve the first target value of the first push index, reducing the risk of the push platform.

[0151] As Figure 5 shown, in another embodiment, a method for processing push messages is provided, which can be applied to the above-mentioned push platform. The method includes the following steps:

[0152] Step 502, receive a competition request sent by the client, where the competition request includes at least one push position.

[0153] The competition request may include the position information corresponding to at least one push position respectively, and the position information can be used to describe the position and / or area size, etc. of the push position in the display page of the client. Optionally, when the competition request includes multiple push positions, the position information may further include the display order corresponding to the push position, etc.

[0154] In some embodiments, the competition request may further include user information and / or search keywords, etc. The user information may include, but is not limited to, user identifiers such as user accounts, ID cards, mobile phone numbers, etc. The search keyword refers to the content entered by the user on the client for searching relevant information. If the competition request is triggered by the user entering a search keyword on the client, the competition request may include the search keyword. Optionally, the push platform may search for user data in the database according to the user information, such as user historical data (such as historical clicked push messages, historical downloaded / purchased push messages, etc.), user habit data (such as user focus), user category (such as gender, age group, product vertical category to which it belongs), etc.

[0155] Step 504, sort the multiple candidate push messages according to the competition information corresponding to the multiple candidate push messages.

[0156] The candidate push information can be push information that is relatively well - matched with the competition request. The multiple candidate push information can be obtained by screening a large amount of push information based on the above - mentioned user data and / or search keywords, etc.

[0157] The push platform can obtain the competition information corresponding to each candidate push information, and calculate the competition score corresponding to each candidate push information according to the competition information corresponding to each candidate push information. The competition score can be used to characterize the competitiveness of the candidate push information. The multiple candidate push information can be sorted in descending order of the competition score.

[0158] As an implementation manner, the revenue per mille corresponding to each candidate push information can be calculated using the above - mentioned formula (1) or formula (2), and the revenue per mille corresponding to the candidate push information can be used as the competition score corresponding to the candidate push information. The multiple candidate push information can be sorted in descending order of the revenue per mille.

[0159] As another implementation manner, based on the score calculation formula, the revenue per mille corresponding to each candidate push information can be calculated according to the competition information corresponding to each of the multiple candidate push information. The score calculation formula can be determined according to the constructed linear programming problem, and the linear programming problem can be used to optimize the target push metrics formulated according to business requirements. The target push metrics can include, but are not limited to, the total overall resource conversion of the push platform (such as the overall GMV of the push platform) or the expected platform revenue of the push platform, etc., but are not limited to this. The competition score corresponding to each candidate push information is calculated based on the score calculation formula, and the competition score can accurately reflect the competitiveness of the push information under the constructed linear programming problem, so as to achieve accurate pushing.

[0160] Step 506: Based on the sorted multiple candidate push information, determine the push information to be pushed corresponding to each push position.

[0161] The candidate push information can be sequentially placed into the push positions in the order from front to back of the sorted multiple candidate push information, so as to obtain the push information to be pushed corresponding to each push position. In some embodiments, each push position may have a corresponding score threshold, and the competition score of the push information to be pushed corresponding to each push position is greater than or equal to the score threshold corresponding to the push position. Taking the first push position as an example, the first push position can be any push position in the competition request, and the competition score of the push information to be pushed corresponding to the first push position is greater than or equal to the score threshold corresponding to the first push position.

[0162] As an implementation manner, in the case of multiple push positions, the multiple push positions can be arranged according to the display order corresponding to each push position to obtain a position queue, and then the sorted multiple candidate push messages are sequentially placed into the arranged multiple push positions. When placing, it can be determined whether the competition score of the current candidate push message is greater than or equal to the score threshold corresponding to the current push position. If the competition score of the current candidate push message is greater than or equal to the score threshold corresponding to the current push position, the current candidate push message is placed into the current push position, the current candidate push message is determined as the push message to be pushed corresponding to the current push position, and the current push position is removed from the position queue. The next candidate push message can be used as the new current candidate push message, and it is sequentially determined whether the remaining push positions in the position queue can be placed.

[0163] If the competition score of the current candidate push message is less than the score threshold corresponding to the current push position, the next push position can be used as the new current push position, and the steps of determining whether the competition score of the current candidate push message is greater than or equal to the score threshold corresponding to the current push position are re-executed until the current push position is the last push position. If the competition scores of the current candidate push messages are all less than the score thresholds corresponding to the push positions, the next candidate push message can be used as the new current candidate push message, and it is sequentially determined whether the remaining push positions in the position queue can be placed.

[0164] Pushing the multiple candidate push messages into the push positions in order from the highest competition score to the lowest, and the competition scores of the push messages to be pushed corresponding to each push position are greater than or equal to the score thresholds corresponding to the push positions, which can ensure the competitiveness of the push messages displayed at each push position and improve the push effect.

[0165] In some embodiments, the score threshold corresponding to the first push position can be determined according to the constraint threshold corresponding to the preset push metric and the deviation information between the first push position and the preset push metric. The preset push metric can be an index for measuring the overall push effect of the push platform. For example, the preset push metric can include the overall estimated click-through rate of the push platform, the overall estimated conversion rate of the push platform, etc., but is not limited thereto. The constraint threshold corresponding to the preset push metric can refer to the constraint of the push platform on the preset push metric, such as the click-through rate threshold corresponding to the overall estimated click-through rate of the push platform, the conversion rate threshold corresponding to the overall estimated conversion rate of the push platform, etc. Optionally, in the constructed linear programming problem, constraint conditions are constructed according to the constraint threshold corresponding to the preset push metric. For example, the constraint condition can be that the preset push metric is greater than or equal to the constraint threshold, so as to ensure that the information push can meet the constraint condition and ensure the overall information push effect and stability of the push platform.

[0166] The deviation information corresponding to the first push position and the preset push metric may refer to the deviation brought by the first push position in the preset push metric. For example, if the preset push metric is the overall estimated click-through rate, the deviation information corresponding to the first push position and the estimated click-through rate may include click deviation, and the expected click-through rate at which the push information is placed in the first push position may be equal to the product of the estimated click-through rate corresponding to the push information and the click deviation corresponding to the first push position.

[0167] Exemplarily, the expected click-through rate ctr of the candidate push information placed in the push position ijk = ctr ij * bias k , where ctr ij represents the estimated click-through rate corresponding to the j-th candidate push information in the i-th competition request, and bias k represents the click deviation corresponding to the k-th push position, and ctr ijk represents the expected click-through rate at which the j-th candidate push information in the i-th competition request is placed in the k-th push position.

[0168] Using the constraint threshold corresponding to the preset push metric and the deviation information corresponding to each push position and the preset push metric, the score threshold corresponding to each push position can be determined, which can more accurately determine the push information to be pushed corresponding to each push position, ensuring the accuracy and push effect of the push.

[0169] Step 508, determine the reserved resource value corresponding to each push information to be pushed according to the resource calculation ratio and the competition information corresponding to each push information to be pushed.

[0170] Step 510, determine the resource value to be deducted corresponding to each push information to be pushed according to the initial resource value and the reserved resource value corresponding to each push information to be pushed.

[0171] Step 512, deduct the numerical resources of each push information to be pushed according to the resource value to be deducted corresponding to each push information to be pushed.

[0172] For the descriptions of steps 508 to 512, reference can be made to the relevant descriptions in the above embodiments, which will not be repeated here.

[0173] In the embodiments of the present application, in the case of receiving a competition request sent by a client, multiple candidate push messages can be sorted according to the competition information respectively corresponding to the multiple candidate push messages, and based on the sorted multiple candidate push messages, the push messages to be pushed corresponding to each push position can be determined, which can improve the accuracy of information push and ensure the overall push effect of the push platform. Moreover, by combining the initial resource value and the reserved resource value of the push message to jointly determine the resource value to be deducted corresponding to the push message, the numerical resources of the push message can be deducted more accurately, and it can be ensured that the push of the push message and the deduction of the numerical resources can achieve the first target value of the first push indicator, improving the stability of the push platform.

[0174] As Figure 6 shown, in one embodiment, the step of sorting multiple candidate push messages according to the competition information respectively corresponding to the multiple candidate push messages may include the following steps:

[0175] Step 602, calculate the competition score corresponding to each candidate push message according to the competition information, the first parameter, and the second parameter respectively corresponding to each candidate push message.

[0176] Among them, the first parameter and the second parameter are determined according to the constructed linear programming problem, and the linear programming problem may include an objective function, a first constraint condition, and a second constraint condition. The objective function, the first constraint condition, and the second constraint condition may be constructed according to different push indicators respectively.

[0177] When the push platform performs information push, it needs to take into account the interests of users, customers, and the push platform at the same time. In the most ideal state, it is necessary to guarantee and optimize the user experience, guarantee and increase the customer revenue, and improve the traffic conversion effect of the push platform. Therefore, it is necessary to model and solve the global objective to obtain the optimal sorting and push mechanism. For example, the estimated click-through rate (CTR) can be used to represent the user experience, ROI can be used to represent the customer interests, and the platform revenue of the push platform can be used to represent the platform interests. Then the business objective is a multi-objective optimization problem and can be described by Equation (10):

[0178]

[0179] If directly modeling according to the above multi-objective optimization problem, the following problems will occur: 1. There are infinitely many sets of Pareto front points for multi-objective optimization, and the problem scale is too large, which will make it difficult to find all the exact Pareto optimal fronts, that is, it is difficult to find the optimal solution; 2. Even if the most accurate Pareto optimal solution can be found, the Pareto optimal solution is not very helpful for the business. From the business perspective, not all Pareto optimal front points will be concerned. For example, the objective with extremely high ROI and extremely low revenue is meaningless for the business.

[0180] Therefore, in the embodiments of the present application, a push metric can be selected as the optimization target, and other push metrics can be constrained. On the premise of ensuring that other push metrics can meet the corresponding constraints, the target can be optimized as much as possible. A linear programming condition can be constructed based on the optimization target and the constraints to be satisfied. The linear programming problem can include an objective function, a first constraint condition, and a second constraint condition, where the objective function, the first constraint condition, and the second constraint condition can respectively correspond to different push metrics.

[0181] In some embodiments, the first constraint condition includes that the first push metric is greater than or equal to the first constraint threshold, the second constraint condition includes that the second push metric is greater than or equal to the second constraint threshold, and the objective function can be used to maximize the third push metric. The first push metric, the second push metric, and the third push metric can all be metrics for measuring the overall push effect of the push platform. Optionally, on the premise of ensuring the user experience and the interests of customers, the traffic monetization efficiency of the push platform can be improved as much as possible, that is, the interests of the push platform can be maximized. Therefore, the first push metric can be a metric for characterizing the interests of customers, the second push metric can be a metric for characterizing the user experience, and the third push metric can be a metric for characterizing the platform interests of the push platform.

[0182] Exemplarily, the first push metric can include the overall ROI of the push platform, the second push metric can include the overall estimated click-through rate or estimated conversion rate of the push platform, etc., and the third push metric can include the overall expected platform revenue or total resource conversion of the push platform, etc., but not limited thereto.

[0183] In some embodiments, the above linear programming problem can be transformed into a dual problem. The first parameter can be the dual variable of the first constraint condition in the dual problem, and the second parameter can be the dual variable of the second constraint condition in the dual problem. The fractional calculation formula can be determined according to the dual problem corresponding to the linear programming problem and the KKT condition. Based on this fractional calculation formula, the competition scores of each candidate push message can be calculated according to the competition information, the first parameter, and the second parameter respectively corresponding to each candidate push message.

[0184] Both the first parameter and the second parameter can be hyperparameters. After constructing the linear programming problem and the dual problem corresponding to the linear programming problem, the first parameter and the second parameter can be optimized to solve the optimal first parameter and second parameter, and then the competition scores of each candidate push message can be determined using the calculated first parameter and second parameter.

[0185] Taking the first push metric as the overall ROI of the push platform, the second push metric as the estimated click-through rate of the push platform, and the third push metric as the expected platform revenue of the push platform as an example, the formalized optimization problem description can be shown as in Equation (11):

[0186]

[0187] Among them, R2 represents the first constraint threshold corresponding to the overall ROI of the push platform, and C represents the second constraint threshold corresponding to the estimated click-through rate of the push platform. The first constraint threshold and the second constraint threshold can be set according to actual needs. The optimization problem described in Equation (11) can be described as a linear programming problem, and this linear programming problem can be shown as in Equation (12):

[0188]

[0189] Among them, ehp k represents the expected exposure probability of the k-th push position, that is, the probability that the push message can be k exposed when placed in the k-th push position; ctr ij represents the estimated click-through rate corresponding to the j-th push message in the i-th competing request; bias k represents the deviation information of the k-th push position, such as click deviation; rev ijk represents the expected platform revenue when the j-th push message is placed in the k-th push position in the i-th competing request, and rev ijk =ehp k *bias k *ctr ij *price ijk , price ijk represents the numerical resources deducted when the j-th push message is placed in the k-th push position in the i-th competing request; gmv ijk represents the expected customer revenue when the j-th push message is placed in the k-th push position in the i-th competing request. The expected customer revenue can be generated successively through steps such as the exposure, click, and conversion of the push message, and gmv ijk =ehp k *bias k *ctr ij *cvr ij *amount ij , cvr ij represents the estimated conversion rate corresponding to the j-th push message in the i-th competing request, and amount ij represents the estimated resource conversion volume corresponding to the j-th push message in the i-th competing request; x ijkis a decision variable, indicating whether the j-th push message in the i-th competing request is placed in the k-th push position. The value can be 0 or 1, where 0 means not placed and 1 means placed, but it is not limited to this.

[0190] The above linear programming problem can be transformed into a dual problem, where the maximization objective function is transformed into the minimization objective function in the dual problem. Since the linear programming problem includes two constraint conditions, the dual problem can include two dual variables (i.e., the first parameter and the second parameter). If the linear programming problem includes one decision variable, the dual problem can include one constraint condition.

[0191] Exemplarily, the dual problem transformed from the linear programming problem in Equation (12) can be as shown in Equation (13):

[0192] min∑ i,j β ij +∑ i,k γ ik

[0193]

[0194] where α1 represents the first parameter, α2 represents the second parameter, α1 is the dual variable of the first constraint condition in the linear programming problem (such as ), and α2 is the dual variable of the second constraint condition in the linear programming problem (such as ).

[0195] Based on the dual problem in Equation (13), when the optimal first parameter α1 and second parameter α2 are given, β ij and γ ik are only related to the i-th competing request and have nothing to do with other competing requests. Therefore, the dual problem in Equation (13) can be further disassembled to obtain Equation (14):

[0196] min∑ i,j β ij +∑ i,k γ ik

[0197] s.t.β ij +γ ik ≥(1 - α1 * R2)rev jk +α1 * gmv jk +α2 * ehp k * bias k * ctr j -α2 * C * ehp k Equation (14);

[0198] The constraint conditions in Equation (14) can be further transformed to obtain Equation (15):

[0199] s.t.β ij +γ ik ≥(1 - α1 * R2)rev jk +α1 * gmv jk +α2 * ehp k *bias k *ctr j -α2 * C * enp k

[0200] =ehp k *(bias k *ctr j *((1 - α1 * R2)*bid j +α1 * cvr j *amount j +α2)-α2 * C) Equation (15);

[0201] where bid j represents the competitive resource value corresponding to the j-th push message.

[0202] Based on the KKT conditions between the linear programming problem and the dual problem, the score calculation formula can be determined. Further, based on the complementary slackness conditions between the linear programming problem and the dual problem, the score calculation formula can be determined. Exemplarily, the complementary slackness conditions between the dual problem and the linear programming problem can be as shown in Equation (16):

[0203]

[0204] In a specific competition request, α2, C, and bias k are all specific values. Therefore, when the result of ctr j *((1 - α1 * R2)*bid j +α1 * cvr j *amount j +α2) is larger, then x ijk =1 is more likely to hold, and the allocation of the push message is closer to the optimal solution. Therefore, the score calculation formula can be as shown in Equation (17):

[0205] score j =ctr j *((1 - α1 * R2)*bid j +α1 * cvr j *amount j +α2) Equation (17).

[0206] In some embodiments, when determining the score calculation formula, based on the score calculation formula, the first parameter may be determined within the first parameter range and the second parameter may be determined within the second parameter range according to the historical competition information set.

[0207] The historical competition information set may include competition information corresponding to multiple historical push messages. The first parameter and the second parameter are parameter combinations in multiple pairs of parameter combinations included in the first parameter range and the second parameter range, which maximize the third push metric when satisfying the first constraint condition and the second constraint condition.

[0208] Furthermore, the historical competition information set may include competition information of at least one historical push message respectively corresponding to multiple historical competition requests.

[0209] The first parameter range refers to the parameter interval to which the first parameter belongs that is preset. The second parameter range may refer to the parameter interval to which the second parameter belongs that is preset. The first parameter range, the second parameter range, and the search step size may be preset. The first parameter may be searched within the first parameter range with the search step size, and the second parameter may be searched within the second parameter range with the search step size. Multiple pairs of parameter combinations are determined from the first parameter range and the second parameter range. Each pair of parameter combinations includes a parameter value corresponding to the first parameter and a parameter value corresponding to the second parameter. The information push effect corresponding to each pair of parameter combinations may be determined according to the historical competition information set, and the information push effects corresponding to different pairs of parameter combinations are compared to find the optimal parameter combination and obtain the optimal first parameter and second parameter.

[0210] As a specific implementation, the current first parameter value α1' can be determined with a search step in the first parameter range [minAlpha1, maxAlpha1], where minAlpha1 represents the lower limit value of the first parameter range and maxAlpha1 represents the upper limit value of the first parameter range. The current second parameter value α2' can be determined with a search step in the second parameter range [minAlpha2, maxAlpha2], where minAlpha2 represents the lower limit value of the second parameter range and maxAlpha2 represents the upper limit value of the second parameter range, to obtain the current parameter combination (α1', α2'). The current parameter combination (α1', α2') can be substituted into the score calculation formula, and based on the competition information corresponding to multiple historical push messages included in the historical competition information set, the simulated competition scores corresponding to each historical push message are calculated, and then the multiple historical push messages are sorted in descending order of the simulated competition scores. Then, based on the sorted multiple historical push messages, the push position allocation and numerical resource deduction are simulated, and then the index values corresponding to the first push index, the second push index, and the third push index of the historical competition information set under the current parameter combination (α1', α2') are determined.

[0211] For example, substitute the current parameter combination (α1', α2') into the score calculation formula (17), and substitute the competition resource values, estimated click-through rates, estimated conversion rates, and estimated resource conversion amounts corresponding to each historical push message included in a historical competition request into the score calculation formula (17) to calculate the simulated competition scores corresponding to each historical push message, and then sort the multiple historical push messages. Based on the sorted multiple historical push messages, the push position allocation of the historical competition request and the numerical resource deduction of the push messages can be simulated. According to the simulation results of each historical competition request, the overall ROI, overall CTR, and expected platform revenue or total resource conversion amount (such as GMV) corresponding to the historical competition information set under the current parameter combination (α1', α2') can be calculated.

[0212] After determining the index values corresponding to the first push index, the second push index, and the third push index respectively for the historical competition information set under the current parameter combination (α1', α2'), the search step can be used to determine the next second parameter value as the new current second parameter value α2' within the second parameter range [minAlpha2, maxAlpha2]. Repeat the above steps until the current second parameter value within the second parameter range is the last parameter value in [minAlpha2, maxAlpha2]. Then, use the search step to determine the next first parameter value as the new current first parameter value α1' within the first parameter range [minAlpha1, maxAlpha1], and continue to repeat the above steps until all parameter combinations included in the first parameter range [minAlpha1, maxAlpha1] and the second parameter range [minAlpha2, maxAlpha2] are traversed, obtaining the index values corresponding to the first push index, the second push index, and the third push index respectively for the historical competition information set under different parameter combinations.

[0213] The index values corresponding to the first push index, the second push index, and the third push index respectively for the historical competition information set under different parameter combinations can be compared, and the parameter combination in which the index value corresponding to the first push index meets the first constraint condition, the index value corresponding to the second push index meets the second constraint condition, and the index value corresponding to the third push index is the largest can be selected as the optimal parameter combination, so as to determine the first parameter and the second parameter. By using the above method to determine the first parameter and the second parameter, it can be ensured that the determined first parameter and second parameter are more accurate, so that the competition scores corresponding to each candidate push information can be calculated more accurately, meeting the goals and corresponding end conditions of global optimization, while taking into account user experience, customer interests, and platform interests, and realizing the globally optimized sorting method.

[0214] In the process of determining the information to be pushed for the current competition request, the pre-set first constraint threshold, second constraint threshold, and the optimal first parameter and second parameter obtained above can be obtained. Based on the score calculation formula, the competition scores corresponding to each candidate push information can be calculated according to the competition information corresponding to each candidate push information, the first constraint threshold, the second constraint threshold, the first parameter, and the second parameter.

[0215] As an implementation manner, the competitive information may include competitive resource values, estimated click-through rates, estimated conversion rates, and estimated resource conversion amounts, etc. The first push metric may include the ROI of the push platform, and the third push metric may include the expected platform revenue or the total amount of resource conversions of the push platform. Taking the first candidate push message among multiple candidate push messages as an example, the first candidate push message is any candidate push message, and based on the score calculation formula, according to the first result, the second result, the second parameter, and the estimated click-through rate corresponding to the first candidate push message, the competitive score corresponding to the first candidate push message can be determined. Among them, the first result can be determined according to the competitive resource value corresponding to the first candidate push message, the first parameter, and the first constraint threshold, and the second result can be determined according to the estimated conversion rate and the estimated resource conversion amount corresponding to the first candidate push message and the first parameter.

[0216] Exemplarily, the first result can be obtained by multiplying the difference between 1 and the first product by the competitive resource value, where the first product is the product of the first parameter and the first constraint threshold; the second result can be the product of the first parameter, the estimated conversion rate, and the estimated resource conversion amount. Taking the score calculation formula (17) as an example, the first result can be (1 - α1 * R2) * bid j and the second result can be α1 * cvr j * amount j .

[0217] It should be noted that the above first push metric, second push metric, and third push metric can be adjusted according to actual business requirements, so the constructed linear programming problem is different. Therefore, the determined score calculation formula will also be different. Exemplarily, assuming that the third push metric is the total amount of resource conversions of the push platform, the optimization problem of the overall objective can be expressed by Equation (18):

[0218]

[0219] Among them, platform GMV represents the total amount of resource conversions of the platform. The optimization problem of Equation (18) can be formulated as a linear programming problem, as shown in Equation (19):

[0220]

[0221] According to the push process similar to the linear programming problem of Equation (12), the score calculation formula can be obtained as shown in Equation (20):

[0222] score j = ctr j * ((1 + α1) * amount j - α1 * R2 * bid j + α2 * ctr j ) Equation (20).

[0223] Taking the fractional calculation formula (20) as an example, the above first result may be α1 * R2 * bid j , and the second result may be (1 + α1) * amount j .

[0224] In the embodiments of the present application, the competition scores corresponding to each candidate push message can be accurately calculated based on the fractional calculation formula. Since this fractional calculation formula is determined based on the linear programming problem, the dual problem of this linear programming problem, and the complementary slackness condition, it has theoretical reliability, and the optimization objective and constraints can be adjusted according to the actual needs of the service, so as to adjust the fractional calculation formula, which has a certain degree of scalability and can be applied to different service scenarios. It can consider the sorting of push messages from the perspective of global optimality, improving the sorting accuracy and the information push effect.

[0225] Step 604: Sort the multiple candidate push messages in descending order of the competition scores.

[0226] After determining the competition scores corresponding to the multiple candidate push messages corresponding to the current competition request, the multiple candidate push messages can be sorted in descending order of the competition scores, and the push positions are placed for the multiple candidate push messages in turn, and the competition scores of the push messages to be pushed corresponding to each push position are greater than or equal to the score threshold corresponding to this push position.

[0227] In some embodiments, taking the first push position in the current competition request as an example, the first push position can be any push position. The score threshold corresponding to the first push position can be determined according to the deviation information corresponding to the first push position and the second push metric, the second parameter, and the second constraint threshold.

[0228] Further, the score threshold corresponding to the first push position can be equal to the second product divided by the deviation information corresponding to the first push position and the second push metric, and the second product can be the product of the second parameter and the second constraint threshold. Exemplarily, taking the second push metric as the estimated click-through rate of the overall push platform, then since β ij +γ ik ≥0, therefore, the competition scores of the candidate push messages need to satisfy the following formula (21):

[0229]

[0230] where bias k represents the deviation information of the kth push position, such as click deviation; C represents the second constraint threshold; α2 represents the second parameter, then the Based on the linear programming problem and its dual problem, the score thresholds corresponding to each push position are determined to make it theoretically reliable, so as to more accurately screen out the push information that meets the requirements for pushing, ensure the competitiveness of the push information displayed at each push position, and guarantee the push effect.

[0231] In the embodiments of the present application, it is possible to start from a global perspective, taking into account the user experience, customer interests, and the interests of the push platform at the same time, realize the ranking of candidate push information, ensure global optimality, improve the accuracy of ranking, and can adjust and optimize the objectives and constraints according to requirements, and can be applied to different business scenarios, with a certain degree of scalability.

[0232] As Figure 7 shown, in one embodiment, another method for processing push information is provided, which can be applied to the above-mentioned push platform. The method may include the following steps:

[0233] Step 710, receive a competition request sent by the client, where the competition request includes at least one push position.

[0234] Step 720, calculate the competition scores corresponding to each candidate push information according to the competition information corresponding to each candidate push information.

[0235] Step 730, sort the multiple candidate push information in descending order of the competition scores.

[0236] Step 740, based on the sorted multiple candidate push information, determine the push information to be pushed corresponding to each push position; wherein, the competition score of the push information to be pushed corresponding to the first push position is greater than or equal to the score threshold corresponding to the first push position, and the first push position is any push position.

[0237] In some embodiments, the score threshold corresponding to the first push position is determined according to the constraint threshold corresponding to the preset push metric and the deviation information corresponding to the first push position and the preset push metric.

[0238] In some embodiments, step 720 includes: calculating the competition scores corresponding to each candidate push information according to the competition information, the first parameter, and the second parameter corresponding to each candidate push information; wherein, the first parameter and the second parameter are determined according to the constructed linear programming problem, and the linear programming problem includes an objective function, a first constraint condition, and a second constraint condition, and the objective function, the first constraint condition, and the second constraint condition are constructed according to different push metrics.

[0239] In some embodiments, the objective function is used to maximize a third push metric, the first constraint condition includes that the first push metric is greater than or equal to a first constraint threshold, and the second constraint condition includes that the second push metric is greater than or equal to a second constraint threshold.

[0240] In some embodiments, the step of calculating the competition score corresponding to each candidate push message according to the competition information, the first parameter, and the second parameter corresponding to each candidate push message includes: calculating the competition score corresponding to each candidate push message based on the score calculation formula according to the competition information, the first parameter, and the second parameter corresponding to each candidate push message.

[0241] The score calculation formula is determined according to the dual problem corresponding to the linear programming problem and the KKT conditions. The first parameter is the dual variable of the first constraint condition in the dual problem, and the second parameter is the dual variable of the second constraint condition in the dual problem.

[0242] In some embodiments, the method further includes: determining the first parameter within the first parameter range and the second parameter within the second parameter range based on the score calculation formula according to the historical competition information set; the historical competition information set includes the competition information corresponding to multiple historical push messages, and the first parameter and the second parameter are the parameter combinations included in the first parameter range and the second parameter range, and are the parameter combinations that maximize the third push metric under the condition of satisfying the first constraint condition and the second constraint condition.

[0243] In some embodiments, the score threshold corresponding to the first push position is determined according to the deviation information corresponding to the first push position and the second push metric, the second parameter, and the second constraint threshold.

[0244] In some embodiments, after determining the push messages to be pushed corresponding to each push position, the method further includes: determining the reserved resource value corresponding to each push message to be pushed according to the resource calculation ratio and the competition information corresponding to each push message to be pushed; the resource calculation ratio is determined according to the first target value of the first push metric; determining the resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message to be pushed; and deducting the numerical resources of each push message to be pushed according to the resource value to be deducted corresponding to each push message to be pushed.

[0245] In some embodiments, the step of determining the reserved resource value corresponding to each push message to be pushed according to the resource calculation ratio and the competition information corresponding to each push message to be pushed includes: determining the target expected revenue according to the resource calculation ratio and the competition information corresponding to each push message to be pushed; determining the reservation ratio according to the target expected revenue; and determining the reserved resource value corresponding to each push message to be pushed according to the reservation ratio and the competition information corresponding to each push message to be pushed.

[0246] In some embodiments, the competition information includes a competition resource value and an estimated display information; the step of determining the target expected revenue according to the resource calculation ratio and the competition information corresponding to each push message to be pushed includes: determining the first expected revenue corresponding to each push message to be pushed according to the resource calculation ratio, the competition resource value corresponding to each push message to be pushed, and the estimated display information; and determining the target expected revenue according to the first expected revenue corresponding to each push message to be pushed.

[0247] In some embodiments, the step of determining the reservation ratio according to the target expected revenue includes: searching for a ratio value in the ratio interval that matches the target expected revenue and using the matching ratio value as the reservation ratio; wherein the estimated expected revenue corresponding to the first ratio value is determined according to the first ratio value and the competition information corresponding to each push message to be pushed, and the first ratio value is any ratio value in the ratio interval.

[0248] In some embodiments, the step of searching for a ratio value in the ratio interval that matches the target expected revenue includes: using the dichotomy method to search for a ratio value in the ratio interval that matches the target expected revenue.

[0249] In some embodiments, the competition information includes a competition resource value and an estimated display information; the step of determining the estimated expected revenue corresponding to the first ratio value according to the first ratio value and the competition information corresponding to each push message to be pushed includes: determining the first resource value corresponding to the first push message according to the first ratio value and the competition resource value corresponding to the first push message; the first push message is any push message to be pushed; determining the second resource value corresponding to the first push message according to the first resource value corresponding to the first push message and the initial resource value; and determining the estimated expected revenue corresponding to the first ratio value according to the second resource value corresponding to each push message to be pushed and the corresponding estimated display information.

[0250] In some embodiments, the resource calculation ratio is determined according to the first target value and historical push data; the historical push data includes the total resource conversion corresponding to the historical time period, the actual display information corresponding to multiple historical push messages pushed in the historical time period, and the deducted numerical resources.

[0251] In some embodiments, the first target value of the first push metric includes the target return on investment, and the actual display information includes the actual click-through rate; the resource calculation ratio is determined according to the first ratio and the actual click-through rates and deducted numerical resources respectively corresponding to a plurality of historical push messages, and the first ratio is the ratio between the total resource conversion amount corresponding to the historical time period and the target return on investment.

[0252] In some embodiments, when there are a plurality of push messages to be pushed, the plurality of push messages are arranged in order; the initial resource value corresponding to the second push message is determined according to the estimated display information of the second push message and the display revenue corresponding to the next push message of the second push message; the second push message is at least one of the plurality of push messages.

[0253] It should be noted that for the description of the method for processing push messages provided in the embodiments of the present application, reference may be made to the relevant descriptions in the above embodiments, and details will not be repeated here.

[0254] In the embodiments of the present application, the competition scores corresponding to each candidate push message may be calculated according to the competition information respectively corresponding to the plurality of candidate push messages, and the plurality of candidate push messages may be sorted in descending order according to the competition scores. Based on the sorted plurality of candidate push messages, the push messages to be pushed corresponding to each push position are determined. Each push position may have a corresponding score threshold, and the competition score of the push message to be pushed corresponding to the push position needs to be greater than or equal to the score threshold corresponding to the push position, so as to ensure the push effect of the push message.

[0255] As Figure 8 shown, in one embodiment, a push message processing device 800 is provided, which can be applied to the above-mentioned push platform. The push message processing device 800 may include an information acquisition module 810, a first resource determination module 820, a second resource determination module 830, and a deduction module 840.

[0256] The information acquisition module 810 is configured to acquire at least one push message to be pushed.

[0257] The first resource determination module 820 is configured to determine the reserved resource value corresponding to each push message according to the resource calculation ratio and the competition information corresponding to each push message; the resource calculation ratio is determined according to the first target value of the first push metric.

[0258] The second resource determination module 830 is configured to determine the resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message.

[0259] The deduction module 840 is configured to deduct the numerical resources of each push message according to the resource value to be deducted corresponding to each push message.

[0260] In one embodiment, the first resource determination module 820 includes a revenue determination unit, a ratio determination unit, and a resource determination unit.

[0261] The revenue determination unit is configured to determine a target expected revenue according to the resource calculation ratio and the competition information respectively corresponding to at least one push message.

[0262] The ratio determination unit is configured to determine a retention ratio according to the target expected revenue.

[0263] The resource determination unit is configured to determine the retained resource value corresponding to each push message according to the retention ratio and the competition information corresponding to each push message.

[0264] In one embodiment, the competition information includes a competition resource value and an estimated display information; the revenue determination unit is further configured to determine a first expected revenue corresponding to each push message according to the resource calculation ratio, the competition resource value corresponding to each push message, and the estimated display information; and determine the target expected revenue according to the first expected revenue corresponding to each push message.

[0265] In one embodiment, the ratio determination unit is further configured to search for a ratio value in the ratio range that matches the estimated expected revenue and the target expected revenue, and use the matching ratio value as the retention ratio; wherein, the estimated expected revenue corresponding to the first ratio value is determined according to the first ratio value and the competition information respectively corresponding to at least one push message, and the first ratio value is any ratio value in the ratio range.

[0266] In one embodiment, the ratio determination unit is further configured to use the dichotomy method to search for a ratio value in the ratio range that matches the estimated expected revenue and the target expected revenue.

[0267] In one embodiment, the competition information includes a competition resource value and an estimated display information; the ratio determination unit is further configured to determine a first resource value corresponding to the first push message according to the first ratio value and the competition resource value corresponding to the first push message; the first push message is any one of the push messages; determine a second resource value corresponding to the first push message according to the first resource value corresponding to the first push message and the initial resource value; and determine the estimated expected revenue corresponding to the first ratio value according to the second resource values respectively corresponding to at least one push message and the corresponding estimated display information.

[0268] In one embodiment, the resource calculation ratio is determined according to the first target value and historical push data; the historical push data includes the total resource conversion amount corresponding to the historical time period, the actual display information respectively corresponding to a plurality of historical push messages pushed in the historical time period, and the deducted numerical resources.

[0269] In one embodiment, the first target value of the first push metric includes the target return on investment, and the actual display information includes the actual click-through rate; the resource calculation ratio is determined according to the first ratio and the actual click-through rates and deducted numerical resources respectively corresponding to a plurality of historical push messages, and the first ratio is the ratio between the total resource conversion amount corresponding to the historical time period and the target return on investment.

[0270] In one embodiment, the information acquisition module 810 includes a request receiving unit, a sorting unit, and a push determination unit.

[0271] The request receiving unit is configured to receive a competition request sent by a client, and the competition request includes at least one push position.

[0272] The sorting unit is configured to sort a plurality of candidate push messages according to the competition information respectively corresponding to the plurality of candidate push messages.

[0273] The push determination unit is configured to determine the push message to be pushed corresponding to each push position based on the sorted plurality of candidate push messages.

[0274] In one embodiment, the sorting unit includes a score calculation subunit and a sorting subunit.

[0275] The score calculation subunit is configured to calculate the competition score corresponding to each candidate push message according to the competition information, the first parameter, and the second parameter respectively corresponding to each candidate push message; wherein, the first parameter and the second parameter are determined according to the constructed linear programming problem, and the linear programming problem includes an objective function, a first constraint condition, and a second constraint condition, and the objective function, the first constraint condition, and the second constraint condition are respectively constructed according to different push metrics.

[0276] The sorting subunit is configured to sort the plurality of candidate push messages in descending order of the competition score.

[0277] In one embodiment, the first constraint condition includes that the first push metric is greater than or equal to the first constraint threshold, and the second constraint condition includes that the second push metric is greater than or equal to the second constraint threshold; the objective function is used to maximize the third push metric.

[0278] In one embodiment, the score calculation subunit is further configured to calculate the competition score corresponding to each candidate push message based on the score calculation formula according to the competition information, the first parameter, and the second parameter respectively corresponding to each candidate push message; the score calculation formula is determined according to the dual problem corresponding to the linear programming problem and the KKT condition, the first parameter is the dual variable of the first constraint condition in the dual problem, and the second parameter is the dual variable of the second constraint condition in the dual problem.

[0279] In one embodiment, the third push metric includes the expected platform revenue or the total resource conversion of the push platform, the first push metric includes the return on investment of the push platform, and the second push metric includes the estimated click-through rate of the push platform; the score calculation subunit is further configured to determine the competition score corresponding to the first candidate push message based on the score calculation formula, the first result, the second result, the second parameter, and the estimated click-through rate corresponding to the first candidate push message; the first candidate push message is any candidate push message, the first result is determined according to the competition resource value corresponding to the first candidate push message, the first parameter, and the first constraint threshold, and the second result is determined according to the estimated conversion rate and the estimated resource conversion amount corresponding to the first candidate push message and the first parameter.

[0280] In one embodiment, the push message processing device 800 further includes a parameter determination module.

[0281] The parameter determination module is configured to determine the first parameter within the first parameter range and the second parameter within the second parameter range based on the score calculation formula according to the historical competition information set; the historical competition information set includes the competition information corresponding to multiple historical push messages, and the first parameter and the second parameter are the parameter combinations included in the first parameter range and the second parameter range, and are the parameter combinations that maximize the third push metric under the condition of satisfying the first constraint condition and the second constraint condition.

[0282] In one embodiment, the competition score of the push message to be pushed at the first push position is greater than or equal to the score threshold corresponding to the first push position, and the first push position is any push position; the score threshold corresponding to the first push position is determined according to the deviation information corresponding to the first push position and the second push metric, the second parameter, and the second constraint threshold.

[0283] In the embodiment of the present application, by combining the initial resource value and the reserved resource value of the push message to jointly determine the resource value to be deducted corresponding to the push message, the numerical resources of the push message can be deducted more accurately, and the reserved resource value is determined according to the resource calculation ratio and the competition information corresponding to the push message, and the resource calculation ratio is determined according to the first target value of the first push metric, so as to ensure that the pushing of the push message and the deduction of the numerical resources can achieve the first target value of the first push metric.

[0284] As Figure 9 shown, in another embodiment, a push message processing device 900 is provided, which can be applied to the above push platform. The push message processing device 900 includes:

[0285] A request receiving module 910, configured to receive a competition request sent by a client, where the competition request includes at least one push position.

[0286] A score calculation module 920 is configured to calculate the competition scores corresponding to each candidate push message according to the competition information corresponding to each of the multiple candidate push messages.

[0287] A sorting module 930 is configured to sort the multiple candidate push messages in descending order of the competition scores.

[0288] A push determination module 940 is configured to determine the push messages to be pushed corresponding to each push position based on the sorted multiple candidate push messages; wherein, the competition score of the push message to be pushed corresponding to the first push position is greater than or equal to the score threshold corresponding to the first push position, and the first push position is any push position.

[0289] In one embodiment, the score threshold corresponding to the first push position is determined according to the constraint threshold corresponding to the preset push metric and the deviation information of the first push position from the preset push metric.

[0290] In one embodiment, the score calculation module 920 is further configured to calculate the competition scores corresponding to each candidate push message according to the competition information, the first parameter, and the second parameter corresponding to each candidate push message; wherein, the first parameter and the second parameter are determined according to the constructed linear programming problem, and the linear programming problem includes an objective function, a first constraint condition, and a second constraint condition, and the objective function, the first constraint condition, and the second constraint condition are constructed according to different push metrics respectively.

[0291] In one embodiment, the objective function is used to maximize the third push metric, the first constraint condition includes that the first push metric is greater than or equal to the first constraint threshold, and the second constraint condition includes that the second push metric is greater than or equal to the second constraint threshold.

[0292] In one embodiment, the score calculation module 920 is further configured to calculate the competition scores corresponding to each candidate push message based on the score calculation formula according to the competition information, the first parameter, and the second parameter corresponding to each candidate push message.

[0293] The score calculation formula is determined according to the dual problem corresponding to the linear programming problem and the KKT conditions. The first parameter is the dual variable of the first constraint condition in the dual problem, and the second parameter is the dual variable of the second constraint condition in the dual problem.

[0294] In one embodiment, the processing device 900 for push messages further includes a parameter determination module.

[0295] A parameter determination module, configured to determine a first parameter within a first parameter range and a second parameter within a second parameter range based on a score calculation formula according to a set of historical competition information; the set of historical competition information includes competition information corresponding to a plurality of historical push messages, and the first parameter and the second parameter are parameter combinations included in the first parameter range and the second parameter range, and are parameter combinations that maximize a third push metric under the condition of satisfying a first constraint condition and a second constraint condition.

[0296] In one embodiment, the score threshold corresponding to the first push position is determined according to the deviation information corresponding to the first push position and the second push metric, the second parameter, and the second constraint threshold.

[0297] In one embodiment, the processing device 900 for push messages further includes a first resource determination module, a second resource determination module, and a deduction module.

[0298] The first resource determination module is configured to determine a reserved resource value corresponding to each push message to be pushed according to a resource calculation ratio and competition information corresponding to each push message to be pushed; the resource calculation ratio is determined according to a first target value of the first push metric.

[0299] The second resource determination module is configured to determine a resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message to be pushed.

[0300] The deduction module is configured to deduct the numerical resources of each push message to be pushed according to the resource value to be deducted corresponding to each push message to be pushed.

[0301] In one embodiment, the first resource determination module includes a revenue determination unit, a ratio determination unit, and a resource determination unit.

[0302] The revenue determination unit is configured to determine a target expected revenue according to the resource calculation ratio and competition information corresponding to each push message to be pushed.

[0303] The ratio determination unit is configured to determine a retention ratio according to the target expected revenue.

[0304] The resource determination unit is configured to determine a reserved resource value corresponding to each push message to be pushed according to the retention ratio and competition information corresponding to each push message to be pushed.

[0305] In one embodiment, the competition information includes a competition resource value and estimated display information; the revenue determination unit is further configured to determine a first expected revenue corresponding to each push message to be pushed according to the resource calculation ratio, the competition resource value corresponding to each push message to be pushed, and the estimated display information; and determine the target expected revenue according to the first expected revenue corresponding to each push message to be pushed.

[0306] In one embodiment, the ratio determination unit is further configured to search for a ratio value in the ratio range where the estimated expected return matches the target expected return, and use the matching ratio value as the retention ratio; wherein, the estimated expected return corresponding to the first ratio value is determined according to the first ratio value and the competition information corresponding to each push message to be pushed, and the first ratio value is any ratio value in the ratio range.

[0307] In one embodiment, the ratio determination unit is further configured to use the dichotomy method to search for a ratio value in the ratio range where the estimated expected return matches the target expected return.

[0308] In one embodiment, the competition information includes a competition resource value and an estimated display information; the ratio determination unit is further configured to determine a first resource value corresponding to the first push message according to the first ratio value and the competition resource value corresponding to the first push message; the first push message is any push message to be pushed; determine a second resource value corresponding to the first push message according to the first resource value corresponding to the first push message and the initial resource value; determine the estimated expected return corresponding to the first ratio value according to the second resource values corresponding to each push message to be pushed and the corresponding estimated display information.

[0309] In one embodiment, the resource calculation ratio is determined according to the first target value and historical push data; the historical push data includes the total resource conversion corresponding to the historical time period, the actual display information corresponding to multiple historical push messages pushed in the historical time period, and the deducted numerical resources.

[0310] In one embodiment, the first target value of the first push metric includes the target return on investment, and the actual display information includes the actual click-through rate; the resource calculation ratio is determined according to the first ratio and the actual click-through rates and deducted numerical resources corresponding to multiple historical push messages respectively, and the first ratio is the ratio between the total resource conversion corresponding to the historical time period and the target return on investment.

[0311] In one embodiment, in the case where there are multiple push messages to be pushed, the initial resource value corresponding to the second push message is determined according to the estimated display information of the second push message and the display return corresponding to the next push message of the second push message; the second expected information is at least one of the multiple push messages to be pushed.

[0312] In the embodiments of the present application, according to the competition information corresponding to multiple candidate push messages respectively, the competition score corresponding to each candidate push message can be calculated, and the multiple candidate push messages can be sorted in descending order of the competition score. Based on the sorted multiple candidate push messages, the push message to be pushed corresponding to each push position can be determined. Each push position can have a corresponding score threshold, and the competition score of the push message to be pushed corresponding to the push position needs to be greater than or equal to the score threshold corresponding to the push position, so as to ensure the push effect of the push message.

[0313] Figure 10 is a block diagram of the structure of an electronic device in an embodiment. As Figure 10 shown, the electronic device 1000 may include one or more of the following components: a processor 1010, and a memory 1020 coupled to the processor 1010. The memory 1020 may store one or more computer programs, and the one or more computer programs may be configured to be executed by one or more processors 1010 to implement the methods described in the above embodiments.

[0314] The processor 1010 may include one or more processing cores. The processor 1010 connects various parts within the entire electronic device 1000 through various interfaces and lines. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 1020, and by calling data stored in the memory 1020, the processor 1010 performs various functions of the electronic device 1000 and processes data. Optionally, the processor 1010 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 1010 may integrate one or a combination of several of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for the rendering and drawing of display content; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor 1010 and may be implemented separately through a communication chip.

[0315] The memory 1020 may include a Random Access Memory (RAM), and may also include a Read-Only Memory (ROM). The memory 1020 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1020 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc. The data storage area may also store data created during the use of the electronic device 1000.

[0316] It can be understood that the electronic device 1000 may include more or fewer structural elements than those in the above structural block diagram. For example, it includes a power module, a WiFi (Wireless Fidelity) module, etc., and no further limitation is provided here.

[0317] An embodiment of the present application discloses a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, it implements the methods described in the above various embodiments.

[0318] An embodiment of the present application discloses a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the methods described in the above various embodiments.

[0319] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it may include the processes of the above-mentioned method embodiments. Among them, the storage medium can be a magnetic disk, an optical disc, a ROM, etc.

[0320] Any reference to memory, storage, database, or other medium as used herein may include non-volatile and / or volatile memory. Suitable non-volatile memory may include ROM, Programmable ROM (PROM), Erasable PROM (EPROM), Electrically Erasable PROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as external cache memory. By way of illustration and not limitation, RAM may be of various forms, such as Static RAM (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus DRAM (RDRAM), and Direct Rambus DRAM (DRDRAM).

[0321] It should be understood that the terms "one embodiment" or "an embodiment" mentioned throughout the specification mean that a particular feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" throughout the specification are not necessarily referring to the same embodiment. Additionally, these particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0322] In various embodiments of the present application, it should be understood that the magnitudes of the sequence numbers of the above processes do not necessarily imply an order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0323] The units described as separate components above may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0324] In addition, each functional unit in the embodiments of the present application may be integrated into one processing unit, may exist physically alone for each unit, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0325] The technical features of the above-described embodiments may be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0326] The above has introduced in detail a method, apparatus, electronic device, and computer program product for processing push information disclosed in the embodiments of the present application. Specific examples are used herein to illustrate the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application. At the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for processing push information, characterized in that, Including: Obtain at least one push message to be pushed; Determine the reserved resource value corresponding to each push message according to the resource calculation ratio and the competition information corresponding to each push message; the resource calculation ratio is determined according to the first target value of the first push metric and historical push data; the first push metric includes the return on investment (ROI) corresponding to the push platform, and the first target value includes the target ROI to be achieved; Determine the resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message; Deduct the numerical resources of each push message according to the resource value to be deducted corresponding to each push message.

2. The method according to claim 1, wherein The step of determining the reserved resource value corresponding to each push message according to the resource calculation ratio and the competition information corresponding to each push message includes: Determine the target expected return according to the resource calculation ratio and the competition information corresponding to each of the at least one push message; Determine the retention ratio according to the target expected return; Determine the reserved resource value corresponding to each push message according to the retention ratio and the competition information corresponding to each push message.

3. The method according to claim 2, wherein The competition information includes the competition resource value and the estimated display information; The step of determining the target expected return according to the resource calculation ratio and the competition information corresponding to each of the at least one push message includes: Determine the first expected return corresponding to each push message according to the resource calculation ratio, the competition resource value corresponding to each push message, and the estimated display information; the first expected return refers to the expected return estimated to be generated by the push platform when pushing the push message; Determine the target expected return according to the first expected return corresponding to each push message, including: accumulating the first expected return corresponding to each push message to obtain the target expected return.

4. The method according to claim 2, wherein The step of determining the retention ratio according to the target expected return includes: Search for a ratio value in the ratio range whose estimated expected return matches the target expected return, and use the matching ratio value as the retention ratio; Among them, the estimated expected return corresponding to the first ratio value is determined according to the first ratio value and the competition information corresponding to each of the at least one push message, and the first ratio value is any ratio value in the ratio range.

5. The method according to claim 4, wherein The competition information includes the competition resource value and the estimated display information; the step of determining the estimated expected return corresponding to the first ratio value according to the first ratio value and the competition information corresponding to each of the at least one push message includes: Determine the first resource value corresponding to the first push message according to the first ratio value and the competition resource value corresponding to the first push message; the first push message is any one of the push messages; Determine the second resource value corresponding to the first push message according to the first resource value and the initial resource value corresponding to the first push message; Determine the estimated expected return corresponding to the first ratio value according to the second resource value corresponding to each of the at least one push message and the corresponding estimated display information.

6. The method according to claim 1, wherein The historical push data includes the total amount of resource conversion corresponding to the historical time period, the actual display information corresponding to multiple historical push messages pushed during the historical time period, and the deducted numerical resources.

7. The method according to claim 6, characterized in that, The first target value of the first push metric includes the target return on investment, and the actual display information includes the actual click-through rate; The resource calculation ratio is determined according to the first ratio, the actual click-through rates corresponding to the multiple historical push messages, and the deducted numerical resources. The first ratio is the ratio between the total amount of resource conversion corresponding to the historical time period and the target return on investment.

8. The method according to any one of claims 1 to 7, characterized in that When there are multiple push messages to be pushed, the multiple push messages are arranged in order; The initial resource value corresponding to the second push message is determined according to the estimated display information of the second push message and the display revenue corresponding to the next push message of the second push message; the second push message is at least one of the multiple push messages.

9. The method according to any one of claims 1 to 7, characterized in that, The obtaining of at least one push message to be pushed includes: Receiving a competition request sent by a client, where the competition request includes at least one push position; Sorting the multiple candidate push messages according to the competition information corresponding to each of the multiple candidate push messages; Based on the sorted multiple candidate push messages, determining the push message to be pushed corresponding to each of the push positions.

10. The method according to claim 9, wherein The sorting of the multiple candidate push messages according to the competition information corresponding to each of the multiple candidate push messages includes: Calculating the competition score corresponding to each of the candidate push messages according to the competition information, the first parameter, and the second parameter corresponding to each of the candidate push messages; where the first parameter and the second parameter are determined according to a constructed linear programming problem, and the linear programming problem includes an objective function, a first constraint condition, and a second constraint condition. The objective function, the first constraint condition, and the second constraint condition are respectively constructed according to different push metrics; when the linear programming problem is transformed into a dual problem, the first parameter is the dual variable of the first constraint condition in the dual problem, and the second parameter is the dual variable of the second constraint condition in the dual problem; Sorting the multiple candidate push messages in descending order of the competition score.

11. The method according to claim 10, characterized in that, The first constraint condition includes that the first push metric is greater than or equal to the first constraint threshold, and the second constraint condition includes that the second push metric is greater than or equal to the second constraint threshold; the objective function is used to maximize the third push metric.

12. The method according to claim 11, wherein The calculating of the competition score corresponding to each of the candidate push messages according to the competition information, the first parameter, and the second parameter corresponding to each of the candidate push messages includes: Calculating the competition score corresponding to each of the candidate push messages based on the score calculation formula according to the competition information, the first parameter, and the second parameter corresponding to each of the candidate push messages; The score calculation formula is determined according to the dual problem corresponding to the linear programming problem and the KKT conditions.

13. The method according to claim 12, wherein The third push metric includes the expected platform revenue or total resource conversion of the push platform, the first push metric includes the return on investment of the push platform, and the second push metric includes the estimated click-through rate of the push platform; the competitive information includes the competitive resource value, estimated click-through rate, estimated conversion rate, and estimated resource conversion volume; Based on the score calculation formula, calculating the competition scores corresponding to the respective candidate push messages according to the competitive information, the first constraint threshold, the first parameter, and the second parameter corresponding to each of the candidate push messages, includes: Based on the score calculation formula, determining the competition score corresponding to the first candidate push message according to the first result, the second result, the second parameter, and the estimated click-through rate corresponding to the first candidate push message; the first candidate push message is any one of the candidate push messages, the first result is determined according to the competitive resource value, the first parameter, and the first constraint threshold corresponding to the first candidate push message, and the second result is determined according to the estimated conversion rate and the estimated resource conversion volume, and the first parameter corresponding to the first candidate push message.

14. The method according to claim 12, characterized in that, Before calculating the competition scores corresponding to the respective candidate push messages according to the competitive information, the first parameter, and the second parameter corresponding to each of the candidate push messages, the method further includes: Based on the score calculation formula, determining the first parameter within the first parameter range and determining the second parameter within the second parameter range according to the historical competition information set; The historical competition information set includes the competitive information corresponding to multiple historical push messages, and the first parameter and the second parameter are the parameter combinations included in the first parameter range and the second parameter range, and are the parameter combinations that maximize the third push metric under the condition of satisfying the first constraint condition and the second constraint condition.

15. The method according to claim 11, wherein The competition score of the push message to be pushed corresponding to the first push position is greater than or equal to the score threshold corresponding to the first push position, and the first push position is any one of the push positions; The score threshold corresponding to the first push position is determined according to the deviation information corresponding to the first push position and the second push metric, the second parameter, and the second constraint threshold.

16. A processing device for pushing information, characterized in that, The device includes: An information acquisition module, configured to acquire at least one push message to be pushed; A first resource determination module, configured to determine the reserved resource value corresponding to each push message according to the resource calculation ratio and the competitive information corresponding to each push message; the resource calculation ratio is determined according to the first target value of the first push metric and historical push data; the first push metric includes the return on investment ROI corresponding to the push platform, and the first target value includes the target ROI to be achieved; A second resource determination module, configured to determine the resource value to be deducted corresponding to each push message according to the initial resource value and the reserved resource value corresponding to each push message; A deduction module, configured to deduct the numerical resources of each push message according to the resource value to be deducted corresponding to each push message.

17. An electronic device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to implement the method according to any one of claims 1 to 15.

18. A computer program product, characterized in that, It includes a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 15 is implemented.

Images