Promotion Information Management Method and Device

Abstract

An embodiment of the present application discloses a cloud-based promotion information management method and device. The method first obtains historical evaluation data related to a promotion initiation object, bins the historical evaluation data to obtain a binned result of the evaluation data of the promotion initiation object, and then processes the binned result of the evaluation data of the promotion initiation object through a trained neural network to obtain an evaluation result. Based on this, the allocation of review resources is performed according to the evaluation result, and the promotion information corresponding to the promotion initiation object is reviewed and managed based on the allocated review resources; since the method first evaluates the promotion initiation object based on historical evaluation data such as the data of the advertisement of the promotion initiation object being rejected, and then schedules cloud review resources according to the evaluation result, it can effectively manage promotion information, that is, it ensures the review and delivery speed of promotion information and the quality of the delivered promotion information.

Description

Technical Field

[0001] This application relates to the field of promotion information management, and particularly to a promotion information management method and device. Background Art

[0002] With the development of network technologies such as communication technologies and interaction technologies, it has become increasingly difficult to control the audience of various promotion information such as advertising information; if the promotion information such as advertising information sent by promotion initiators such as advertisers carries negative information such as illegal publicity or illegal information, these promotion information will cause extremely bad social impacts after being put on the network.

[0003] In order to ensure that the promoted information put on the market has no negative information, promotion information operators such as Internet advertising operators need to conduct manual review and manual inspection of the promotion information before and during the promotion information release to ensure that the promoted information put on the market has no illegal publicity or malicious tampering; with the increasing number of promotion initiators such as advertisers, the review resources are limited and increasingly insufficient, resulting in the inability to effectively review the promotion information.

[0004] That is, the current promotion information management technology at least has technical problems such as the inability to effectively manage the promotion information.

[0005] Application Content

[0006] The embodiments of this application provide a promotion information management method and device to alleviate the technical problem that the current promotion information management technology cannot effectively manage the promotion information.

[0007] To solve the above technical problems, the embodiments of this application provide the following technical solutions:

[0008] The embodiments of this application provide a promotion information management method, which includes:

[0009] Training data parameters and binning parameters according to the training data corresponding to multiple initial data parameters of the sample object;

[0010] Obtaining historical evaluation data related to the promotion initiator according to the data parameters;

[0011] Binning the historical evaluation data according to the binning parameters to obtain the evaluation data binning result of the promotion initiator, and obtaining the evaluation result of the promotion initiator according to the evaluation data binning result of the promotion initiator;

[0012] Allocating corresponding promotion information review resources to the promotion initiator according to the evaluation result;

[0013] Using the promotion information review resources to review and manage the promotion information corresponding to the promotion initiator.

[0014] An embodiment of the present application provides a promotion information management device, which includes:

[0015] A parameter training module, configured to train data parameters and binning parameters according to training data corresponding to multiple initial data parameters of sample objects;

[0016] A data acquisition module, configured to acquire historical evaluation data related to a promotion initiation object according to the data parameters;

[0017] An object evaluation module, configured to bin the historical evaluation data according to the binning parameters to obtain a binning result of the evaluation data of the promotion initiation object, and obtain an evaluation result of the promotion initiation object according to the binning result of the evaluation data of the promotion initiation object;

[0018] A resource allocation module, configured to allocate corresponding promotion information review resources to the promotion initiation object according to the evaluation result;

[0019] An information management module, configured to review and manage the promotion information corresponding to the promotion initiation object by using the promotion information review resources.

[0020] An embodiment of the present application provides a computer device, which includes a processor and a memory. The memory stores multiple instructions, and the instructions are suitable for being loaded by the processor to execute the steps in the above method.

[0021] An embodiment of the present application provides a computer-readable storage medium. The computer-readable storage medium stores multiple instructions, and the instructions are suitable for being loaded by the processor to execute the steps in the above method.

[0022] An embodiment of the present application provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the above method.

[0023] The embodiments of the present application provide a promotion information management method and device. The method first obtains historical evaluation data related to the promotion initiator, bins the historical evaluation data to obtain the evaluation data binning result of the promotion initiator, and then processes the evaluation data binning result of the promotion initiator through a trained neural network to obtain the evaluation result of the promotion initiator. On this basis, the audit resources are allocated according to the evaluation result, and the promotion information corresponding to the promotion initiator is audited and managed based on the allocated audit resources. Since this method first evaluates the promotion initiator based on the neural network according to historical evaluation data such as the advertisement rejection data of the promotion initiator such as the advertiser, and obtains the evaluation result of whether the promotion initiator such as the advertiser is likely to initiate the intention of promoting information with negative information, and then allocates the audit resources according to this evaluation result to audit and manage the promotion information corresponding to the promotion initiator such as the advertiser. For example, if the evaluation result of a certain advertiser indicates that it is very likely to initiate and release an advertisement including negative information, then all advertisements initiated by this advertiser are subject to key audits (usually by manual audits) and management. For other advertisers, the evaluation result indicates that it is very unlikely to initiate and release an advertisement including negative information, then all advertisements initiated by this advertiser are subject to general audits (usually by machine audits) and management. In this way, a reasonable audit resource scheduling can be realized based on the promotion initiator such as the advertiser to manage the promotion information. For example, the audit resources are scheduled based on cloud technology, so that the audit resources can be effectively allocated to the promotion information that needs to be audited, and then the promotion information can be effectively managed, alleviating the technical problem that the current promotion information management technology cannot effectively manage the promotion information, that is, ensuring the audit and release speed of the promotion information and the quality of the released promotion information. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0025] Figure 1 It is a networking schematic diagram of the promotion system provided by the embodiments of the present application.

[0026] Figure 2 It is the first process schematic diagram of the promotion information management method provided by the embodiments of the present application.

[0027] Figure 3 It is the second process schematic diagram of the promotion information management method provided by the embodiments of the present application.

[0028] Figure 4It is the third process schematic diagram of the promotion information management method provided by the embodiments of the present application.

[0029] Figure 5 It is the structural schematic diagram of the promotion information management device provided by the embodiments of the present application.

[0030] Figure 6 It is the structural schematic diagram of the computer device provided by the embodiments of the present application.

[0031] Figure 7 It is the schematic diagram of the data type involved in the embodiments of the present application.

[0032] Figures 8a to 8d It is the interface schematic diagram involved in the embodiments of the present application. Detailed implementation manners

[0033] 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 skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0034] The promotion information management method involved in the embodiments of the present application can be implemented through a server, specifically through a server in a cloud system. For example, the scheduling of audit resources can be based on cloud technology to ensure the effective utilization of audit resources.

[0035] Cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or a local area network to achieve data calculation, storage, processing, and sharing. Cloud technology is the general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model, which can form a resource pool, be used on demand, and be flexible and convenient. Cloud computing technology will become an important support. The background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites, and more portal websites. With the high development and application of the Internet industry, in the future, each item may have its own identification mark and needs to be transmitted to the background system for logical processing. Data at different levels will be processed separately, and various industry data requires a powerful system support, which can only be achieved through cloud computing. In the present application, cloud technology can be used to implement the audit function corresponding to audit resources. For example, based on a neural network, it is determined whether there is illegal information, illegal data, etc. in the audit promotion information to achieve the audit of promotion information. The present application mainly performs automated allocation of cloud audit resources based on the evaluation results to improve the rationality of resource allocation.

[0036] Please refer toFigure 1 , Figure 1 This is a schematic diagram of the scenario of the promotion system provided by the embodiments of the present application. The system may include user-side devices and service-side devices. The user-side devices and the service-side devices are connected through the Internet composed of various gateways, etc., which will not be elaborated here. Among them, the user-side devices include multiple terminals 11, and the service-side devices include multiple servers 12; wherein:

[0037] The terminal 11 includes but is not limited to portable terminals such as mobile phones and tablets, as well as fixed terminals such as computers, inquiry machines, and advertising machines. It is a service port that users can use and operate. In the present application, the terminal can provide an audit window for auditors, etc., and can also provide a promotion window for promotion initiators such as advertisers, etc.; for the convenience of the following description, the terminal 11 is defined as an audit terminal 11a and a promotion terminal 11b. The audit terminal 11a is used to specify sample users, display promotion information to be audited, etc., while the promotion terminal 11b is used to upload promotion information and display the delivery results, etc.;

[0038] The server 12 provides various business services for users, including a promotion server 12a, an evaluation server 12b, etc. Among them, the evaluation server 12b is used for services such as model training and evaluating promotion initiators such as advertisers, etc. The promotion server 12a is used to receive promotion information from the promotion terminal, call audit resources for auditing, and conduct promotions according to the audit results.

[0039] In this application, the evaluation server 12b is used to obtain historical evaluation data related to the promotion initiation object according to the trained data parameters; bin the historical evaluation data according to the trained binning parameters to obtain the evaluation data binning result of the promotion initiation object; and process the evaluation data binning result of the promotion initiation object through the trained neural network to obtain the evaluation result of the promotion initiation object. The promotion server 12a can be used to allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result, and use the promotion information review resources to review and manage the promotion information corresponding to the promotion initiation object. For example, if the evaluation result of a certain advertiser indicates that it is very likely to initiate and launch an advertisement including negative information, then all advertisements initiated by this advertiser are subject to key review (generally by manual review) management. For some other advertisers whose evaluation results indicate that it is very unlikely to initiate and launch an advertisement including negative information, then all advertisements initiated by this advertiser are subject to general review (generally by machine review) management. In this way, reasonable review resource scheduling can be realized based on promotion initiation objects such as advertisers to manage promotion information. For example, review resource scheduling is performed based on cloud technology, so that review resources can be effectively allocated to the promotion information that needs to be reviewed, and then promotion information can be effectively managed, alleviating the technical problem that the current promotion information management technology cannot effectively manage promotion information, that is, ensuring the review and delivery speed of promotion information and the quality of the delivered promotion information.

[0040] In this application, the promotion server 12a and the evaluation server 12b can be independent physical servers, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, and this application does not make any restrictions here.

[0041] In this application, the evaluation server 12b can be used as a node in the blockchain, so that the evaluation results of each promotion initiation object such as each advertiser can be traced, ensuring data security.

[0042] It should be noted that Figure 1The schematic diagram of the system scenario shown is only an example. The servers and scenarios described in the embodiments of this application are used to more clearly illustrate the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those of ordinary skill in the art can know that with the evolution of the system and the emergence of new business scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0043] Figure 2 is the first process schematic diagram of the promotion information management method provided by the embodiments of this application. This method can be applied in one or more servers; please refer to Figure 2 and this promotion information management method includes the following steps:

[0044] 201: Train data parameters and binning parameters according to the training data corresponding to multiple initial data parameters of the sample object.

[0045] In one embodiment, the data parameters in this application include data types and time parameters. The data types can be penalty histories, opening durations, etc. under account attributes. The time parameters can be penalty histories in the last 3 days, penalty histories in the last 15 days, etc.; the binning parameters in this application can be equal division parameters, equal depth parameters, etc. For example, for the data under a certain time parameter dimension of the data type of penalty history in the last 3 days, the penalty history includes 0 times, 100 times, etc., and the binning parameters can include [0, 10], [11, 20], etc. It is also possible to perform equal depth binning on advertisers according to the number of advertisers corresponding to each number value (such as 0 times, 100 times, etc.). For example, for 10,000 advertisers, each advertiser has a different number of penalty times. According to the penalty times, the advertisers can be assigned to the corresponding bins. For example, all advertisers with penalty times from 0 to 10 times are assigned to the bin corresponding to the binning parameter [0, 10]. If the number of advertisers corresponding to a certain penalty time is relatively large, this number can be used as a binning parameter. For example, if the number of advertisers with a penalty time of 25 times is greater than 1000, then 25 is used as a binning parameter

[25] . Equal depth binning can be achieved by adjusting the binning parameters so that the total number of advertisers in each bin is approximately the same. To ensure the accuracy of the evaluation results, it is necessary to consider as much data as possible in different dimensions and as comprehensive data as possible; however, some data has no impact on the evaluation results, and some data only requires obtaining data in the recent few days. Currently, it is required that the back-end personnel determine which data and which time periods of data to select based on experience to determine the data parameters, which requires high requirements for the back-end personnel; similarly, the binning parameters are also determined by the back-end personnel based on experience, and it is difficult to determine the rationality.

[0046] Based on this, the present application provides a mechanism that can train the acquired parameters and binning parameters to avoid the errors of manual experience. In one embodiment, this step includes: acquiring a plurality of initial data parameters and a plurality of initial binning parameters; acquiring the evaluation data corresponding to the plurality of initial data parameters of the sample users and determining it as training data; training the plurality of initial data parameters and the plurality of initial binning parameters according to the training data to obtain the trained data parameters and the trained binning parameters; the trained data parameters belong to the plurality of initial data parameters.

[0047] In one embodiment, the step of acquiring the trained data parameters and the trained binning parameters includes: acquiring a plurality of initial data types, a plurality of initial time parameters corresponding to each initial data type, and a plurality of initial binning parameters corresponding to each initial data type; acquiring the first identifier of the first sample object corresponding to the first evaluation result and the second identifier of the second sample object corresponding to the second evaluation result, where the first evaluation result and the second evaluation result are different; acquiring the training data according to the first identifier, the second identifier, the plurality of initial data types, and the plurality of initial time parameters corresponding to each initial data type; performing a first data parameter screening on the plurality of initial data types and the plurality of initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; performing a first training data screening on the training data according to the screening result of the first data parameter screening; performing binning processing on the screening result of the first training data screening according to the plurality of initial binning parameters corresponding to each initial data type; performing binning parameter screening on the plurality of initial binning parameters according to the binning results corresponding to each initial binning parameter to obtain the trained binning parameters; acquiring the index parameters of the binning results of the trained binning parameters; performing a second data parameter screening on the plurality of initial data types and the plurality of initial time parameters corresponding to each initial data type according to the index parameters to obtain the trained data parameters. In some embodiments of the present application, if the evaluation result of an advertiser indicates that the advertiser is a good advertiser (marked as 0) or a bad advertiser (marked as 1), then the first evaluation result and the second evaluation result can be a good advertiser (marked as 0) or a bad advertiser (marked as 1) respectively. Correspondingly, the first sample object is the sample object corresponding to a good advertiser (marked as 0), and the second sample object is the sample object corresponding to a bad advertiser (marked as 1). The identifier of the object can include unique identification codes such as the advertiser's account and mobile phone number.

[0048] For example, the present application may first set 100 initial data types at different angles, set 3 or 9 initial time parameters for each initial data type, set 4 or 6 initial binning parameters for each initial data type, and then determine the data corresponding to these initial data types of a predetermined number of sample objects (such as 100,000 positive sample objects and 100,000 negative sample objects) based on the marks of the background users as training data. After the first data parameter screening of the 100 initial data types according to the training data, only 50 data types are left, and only 1 or 2 time parameters of these 50 data types are retained; for the data of each time parameter under each data type, different initial binning parameters are used, and the binning results are different. The present application screens the initial binning parameters according to the binning results to obtain the final binning parameters. For example, for the data of each time parameter under each data type, only one binning parameter is retained. Then, for the binning results of the data of each time parameter under each data type, the index parameters of the binning results are obtained. According to these index parameters of the binning results, the second screening of the data parameters is performed, which may include 10 data types, and these 10 data types also only include 1 time parameter. Different data types may correspond to different time parameters or different binning parameters.

[0049] In one embodiment, the step of performing the first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data includes: determining the data missing rate of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; and performing the first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the data missing rate of each initial time parameter under each initial data type. For example, if the initial data type includes the data type of Maple Leaf complaint (Maple Leaf is a complaint web page), and if 5% of the sample objects have this data and the remaining 95% of the sample objects do not have this data, the data missing rate of all initial time parameters under this initial data type is 95%, then this data type is deleted.

[0050] In one embodiment, the step of performing a first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data includes: determining the data single-value proportion of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; and performing a first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the data single-value proportion of each initial time parameter under each initial data type. For example, if the initial data type includes the data type of abnormal change, and if the abnormal changes of all objects occur in the most recent 10 days and not in the most recent 3 days, that is, under this data type, the data of the time parameter (the most recent 3 days) are all 0 and the data single-value proportion is 100%, then the time parameter (the most recent 3 days) under this data type is deleted.

[0051] In one embodiment, the step of performing a binning parameter screening on the multiple initial binning parameters according to the binning results corresponding to each initial binning parameter to obtain the trained binning parameters includes: determining the data distribution information of each initial time parameter of each initial data type corresponding to each initial binning parameter according to the binning results corresponding to each initial binning parameter; and performing a binning parameter screening on the multiple initial binning parameters corresponding to each initial data type according to the data distribution information. The data distribution information may include the proportion of objects in each bin. For example, for the data type of opening time, based on different binning parameters, 200,000 data are divided into 10 bins. If, after binning based on the binning parameter a, the number of objects in one or several bins is significantly less than that in other bins, then the binning parameter a is deleted. If, after binning based on the binning parameter b, the number of objects in each bin is the same or approximately the same, then the binning parameter b is retained.

[0052] In one embodiment, in order to ensure the binning effect, different from the existing method that can only use the data range as the binning parameter, the binning parameters in this application can include the data range or a data point. For example, the binning parameter can be [1, 10] or

[11] , that is, the objects whose data belong to the data range [1, 10] are put into one bin, and the objects whose data is

[11] are put into another bin, so as to ensure the uniformity of the binning result as much as possible.

[0053] In one embodiment, the metric parameter includes an information quantity parameter. The step of performing a second data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the metric parameter to obtain the trained data parameters includes: obtaining the basic variables corresponding to each initial data type; obtaining the information quantity parameters of the training data corresponding to at least two different initial data types for which the basic variables meet the preset conditions and the multiple initial time parameters corresponding to each initial data type; and performing a second data parameter screening on the at least two different initial data types that meet the preset conditions and the multiple initial time parameters corresponding to each initial data type according to the information quantity parameter to obtain the trained data parameters. For example, for the originality and quality score in the initial data type, their basic variable is the official account quality, and one of them can be selected according to the information quantity IV parameter corresponding to the binning result to complete the second screening of the data parameters.

[0054] In one embodiment, the metric parameter includes an information quantity parameter and a data type correlation coefficient. The step of performing a second data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the metric parameter to obtain the trained data parameters includes: obtaining the information quantity parameters of the training data corresponding to each initial data type and the multiple initial time parameters corresponding to each initial data type; obtaining the data type correlation coefficient between each initial data type; and performing a second data parameter screening on each initial data type and the multiple initial time parameters corresponding to each initial data type according to the information quantity parameter to obtain the trained data parameters. For example, for the originality and quality score in the initial data type, their data type correlation coefficient (such as the Pearson correlation coefficient) r indicates that these two data types are highly correlated, and one of them can be selected according to the information quantity IV parameter corresponding to the binning result to complete the second screening of the data parameters.

[0055] In one embodiment, the neural network includes a classification network and a conversion network; the step of obtaining the trained neural network includes: processing the training data according to the trained data parameters and the trained binning parameters to obtain effective training data; training the classification network with the effective training data to obtain the trained classification network; training the conversion network according to the display parameters corresponding to the evaluation results to obtain the trained conversion network; and determining the trained neural network according to the trained classification network and the trained conversion network. The classification network can be any binary classification neural network, such as a GBDT neural network, a regression neural network, etc. Hereinafter, a regression neural network will be taken as an example for detailed description. Since the result output by the classification network is the probability that the object is a good object or a bad object, and its value range is [0, 1], it is not convenient for users to understand and is not convenient for users to set the management mode. Then, the conversion network can convert the result output by the classification network into a score value, and the representational meaning of the score value (for example, the higher the score, the worse the object, or the lower the score, the worse the object) can be set as needed. The conversion network can be any neural network with data conversion function. Hereinafter, a proportional logarithmic function will be taken as an example for description.

[0056] 202: Obtain historical evaluation data related to the promotion initiation object according to the data parameters.

[0057] In one embodiment, for example, the trained data parameters include 10 different types of data and the corresponding time parameters for each data type. At this time, according to the promotion initiation object, such as the user identifier of the advertiser, the data of these 10 data types at the corresponding time parameters can be obtained from one or more paths as the historical evaluation data related to the promotion initiation object for the operation of the next step.

[0058] 203: Bin the historical evaluation data according to the binning parameters to obtain the binning result of the evaluation data of the promotion initiation object, and obtain the evaluation result of the promotion initiation object according to the binning result of the evaluation data of the promotion initiation object.

[0059] In one embodiment, a binning parameter corresponds to each data type. Based on this binning parameter, all the data under this data type fall into one bin; after the training is completed, each bin corresponds to a weight of evidence WOE. In this way, according to the bin into which a certain data type in the historical evaluation data falls, the weight of evidence WOE corresponding to this data type can be determined. After all the data of all data types of this object are binned, the weights of evidence WOE corresponding to all data types can be obtained, that is, the binning result of the evaluation data of the promotion initiation object is obtained. For example, the binning result of the evaluation data of an object is (x1, x2,..., x10), where x1 = 0.3, x2 = 0.09, etc.

[0060] In one embodiment, the step of obtaining the evaluation result of the promotion initiation object according to the binned result of the evaluation data of the promotion initiation object may include: obtaining a trained neural network; using the trained neural network to process the binned result of the evaluation data of the promotion initiation object to obtain the evaluation result of the promotion initiation object. Based on the neural network, a better evaluation result can be obtained.

[0061] In one embodiment, the binned result of the evaluation data of a certain object can be directly substituted into a classification network to obtain the probability p that the object is a bad object or the probability p that the object is a good object, and then p is substituted into a conversion network to obtain the corresponding score of the object, such as 80 or 20, etc., as the evaluation result of the object.

[0062] 204: Allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result.

[0063] In one embodiment, the promotion information review resources may include at least one of manual review resources and machine review resources.

[0064] In one embodiment, two different management methods can be simply set, such as two review methods of manual review and machine review. At this time, this step includes: determining the promotion information review method corresponding to the promotion initiation object according to the association relationship between the evaluation result and the review mode and the evaluation result of the promotion initiation object; using the promotion information review method corresponding to the promotion initiation object to allocate corresponding promotion information review resources for the promotion initiation object. In this embodiment, objects with scores greater than a threshold (such as 50) can be allocated manual review resources, and objects with scores less than the threshold (such as 50) can be allocated machine review resources.

[0065] In one embodiment, different review resources can be allocated to objects with different score levels. For example, this step includes: determining the promotion information review resources corresponding to the promotion initiation object according to the corresponding relationship between the evaluation result and the proportion of review resources and the evaluation result of the promotion initiation object. For example, 50% of the review resources are allocated to objects with scores in [81, 90], and all promotion information of objects with this score value is manually reviewed. 20% of the review resources are allocated to objects with scores in [51, 60], and 20% of the promotion information of objects with this score value is randomly selected for manual review, etc.

[0066] 205: Use the promotion information review resources to conduct review management on the promotion information corresponding to the promotion initiation object.

[0067] In one embodiment, the review includes a preliminary review when placing an advertisement and a regular review during the advertisement placement period, etc. The promotion information corresponding to the promotion initiation object may include the promotion information initiated by the promotion initiation object, the forwarded promotion information, the followed promotion information, or the collected promotion information, etc.

[0068] In one embodiment, in this embodiment, the promotion information corresponding to the objects with scores greater than the threshold (for example, 50) can be all manually reviewed, and the promotion information corresponding to the objects with scores less than the threshold (for example, 50) can be all machine-reviewed. Or for example, allocate 50% of the review resources to the objects with scores in the range of [81, 90], and conduct all manual reviews on the promotion information of the objects with this score value. Allocate 20% of the review resources to the objects with scores in the range of [51, 60], and randomly select 20% of the promotion information of the objects with this score value for manual review, etc.

[0069] In one embodiment, the review personnel can be reminded of the key application promotion information. Figure 2 After the step of obtaining the evaluation result of the promotion initiation object in the method shown, it further includes: obtaining an alarm threshold; determining the target promotion initiation object that needs to be alarmed according to the evaluation result of the promotion initiation object and the alarm threshold; sending an alarm message to the management device, and the alarm message carries the object information of the target promotion initiation object. For example, if the alarm threshold is 60, then the score of a certain object is 55, so it is not used as the target promotion initiation object. If the score of a certain object is 75, then it is used as the target promotion initiation object, and then an alarm message is sent to the management device.

[0070] In one embodiment, the backstage personnel can actively view the scores of certain objects, or the score distribution of the active objects within a certain time period. At this time Figure 2 After the step of obtaining the evaluation result of the promotion initiation object in the method shown, it further includes: receiving an evaluation result query request sent by the management device, and the evaluation result query request carries the object identifier of the promotion initiation object; obtaining the evaluation result of the promotion initiation object corresponding to the object identifier and the evaluation result-related description information according to the object identifier; sending the evaluation result of the promotion initiation object and the evaluation result-related description information to the management device, so that the management device displays the evaluation result of the promotion initiation object and the evaluation result-related description information on the information display interface. On this basis, the management terminal can display the score of a single object and the main basis for scoring, or can also display the overall distribution situation, etc., which will be described in detail below.

[0071] In one embodiment, if the backstage personnel analyze and determine that the score of a certain object is incorrect, they can also modify and feedback these incorrect scores. At this time Figure 2After the step of sending the evaluation result of the promotion initiation object and the description information related to the evaluation result to the management device, the method further includes: receiving the feedback information sent by the management device; correcting the evaluation result of the promotion initiation object according to the feedback information to obtain a corrected evaluation result; and managing the promotion information corresponding to the promotion initiation object according to the corrected evaluation result. For example, if the server determines that the score of a certain object is 80, but the back-end personnel think that some data of the object are incorrect and modify its score to 20, then the server modifies the corresponding score according to the feedback from the back-end personnel and manages the promotion information of the object according to the modified score.

[0072] This embodiment provides a method for managing promotion information. First, based on a neural network, the promotion initiation object is evaluated according to historical evaluation data such as the advertisement rejection data of the promotion initiation object such as the advertiser, and an evaluation result of whether the promotion initiation object such as the advertiser is likely to initiate the intention of promoting information with negative information is obtained. Then, according to this evaluation result, the promotion information corresponding to the promotion initiation object such as the advertiser is managed. For example, if the evaluation result of a certain advertiser indicates that it is very likely to initiate and release an advertisement including negative information, then all advertisements initiated by this advertiser are subject to key review (generally manual review) management. For some other advertisers, if the evaluation result indicates that it is very unlikely to initiate and release an advertisement including negative information, then all advertisements initiated by this advertiser are subject to general review (generally machine review) management. In this way, reasonable review resource scheduling can be realized based on the promotion initiation object such as the advertiser to manage promotion information. For example, based on cloud technology for review resource scheduling, so that the review resources can be effectively allocated to the promotion information that needs to be reviewed, and then the promotion information can be effectively managed, alleviating the technical problem that the current promotion information management technology cannot effectively manage promotion information, that is, ensuring the review and release speed of promotion information and ensuring the quality of the released promotion information.

[0073] In this application, the promotion information may include advertisements, microblogs, news bulletins, etc. Correspondingly, the promotion initiation object may be an advertiser, a microblog account, a news account, etc.; for different promotion initiation objects, the corresponding data types, time parameters, binning parameters, and neural networks may be different, but their training methods, calculation methods, and management method types Figure 3 and Figure 4 The illustrated embodiment further describes the present application with an advertiser as a specific scenario, and the implementation of other scenarios is similar.

[0074] Figure 3 The illustrated scenario focuses on how to manage the advertisements of advertisers. Figure 4The scenario shown focuses on how to display the scores of some or all advertisers, risk warnings, etc. In this embodiment, for the selection of sample advertisers, advertisers who have launched advertisements containing negative information are marked as bad samples, and advertisers who have not launched advertisements containing negative information are marked as good samples. For a certain advertiser, the higher the score output by the neural network, the greater the possibility that the advertiser is a bad advertiser (there is a risk of sending advertisements containing negative information), and the advertisements of this advertiser need to be key managed.

[0075] Figure 3 This is the second process schematic diagram of the promotion information management method provided by the embodiments of the present application. Please refer to Figure 3 , and the promotion information management method includes the following steps:

[0076] Steps 301 to 303: Perform training on data parameters, binning parameters, and neural networks.

[0077] In this step, the back-end personnel use the audit terminal to set sample advertisers, initial data types, initial time parameters, initial binning parameters, and neural networks, and then trigger the evaluation server to perform parameter and neural network training.

[0078] In this embodiment, according to whether each advertiser in the historical data has launched advertisements containing negative information such as illegal marketing, advertisers who have launched advertisements containing negative information are marked as bad advertisers, and advertisers who have not launched advertisements containing negative information are marked as good advertisers. Then, 10,000 good advertisers (i.e., the second sample object in the above text) and 10,000 bad advertisers (i.e., the first sample object in the above text) are randomly selected from the already marked advertisers as sample objects. Based on the account identifiers of these advertisers, all sample evaluation data can be obtained.

[0079] Since before training, the back-end personnel do not know which data types are effective for evaluating the quality of advertisers and the influence degree of effective data types on the evaluation results, in this embodiment, when setting the initial data type, as Figure 7 shown, the data types for advertiser evaluation data can include data from different perspectives. For example:

[0080] The opening time C1 related to the account attribute (i.e., the duration from the advertiser's opening time to the current time, such as 180 days, 3 days, etc., or whether the advertiser is opened within a predetermined duration), the penalty history C2 (i.e., the number of times the advertiser is penalized for sending negative information or other operations within a predetermined duration, such as 0 penalties, 10 penalties, etc.), the consumption history C3 (i.e., the number of advertisement information launched by the advertiser within a predetermined duration, such as 3 launches, 10 launches, etc.), the endorsement situation C4 (i.e., the number of advertising merchants served by the advertiser within a predetermined duration, such as 2 merchants, 8 merchants, etc.), etc.

[0081] Abnormal changes related to operation behavior C5 (i.e., the number of times the advertiser's name, entity, etc. are changed within a predetermined time period, such as 0 times, 3 times, etc.), abnormal logins C6 (i.e., the number of times the advertiser logs in outside the place of habitual residence within a predetermined time period, such as 0 times, 3 times, etc.);

[0082] Initial review rejections C7 related to review behavior (i.e., the number of times the advertisement review initiated by the advertiser is directly not approved within a predetermined time period, such as 0 times, 3 times, etc.), inspection rejections C8 (i.e., the number of times the advertisement initiated by the advertiser is determined to contain negative information and rejected during the recheck within a predetermined time period, such as 0 times, 3 times, etc.);

[0083] Complaints related to complaint information: Own - entrance complaints C9 (i.e., the number of times the advertiser is complained by users through the complaint window provided by the advertising operator within a predetermined time period, such as 10 times, 30 times, etc.), official - account complaints C10 (i.e., the number of times the advertiser is complained by users through the official account of the advertising operator within a predetermined time period, such as 10 times, 30 times, etc.), Fengye - page complaints C11 (i.e., the number of times the advertiser is complained by users through the Fengye website within a predetermined time period, such as 10 times, 30 times, etc.), instant - messaging complaints C12 (i.e., the number of times the advertiser is complained by users through instant - messaging applications within a predetermined time period, such as 10 times, 30 times, etc.);

[0084] Regarding the quality of the official account: Opening time C13 (i.e., the duration from the opening time of the official account bound by the advertiser to the current time, such as 180 days, 3 days, etc., or whether the official account bound by the advertiser is opened within a predetermined time period), penalty history C14 (i.e., the number of times the advertiser's official account is penalized for posting negative information or other operations within a predetermined time period, such as 0 penalties, 10 penalties, etc.), originality C15 (i.e., the ratio of original articles among the articles published by the advertiser's official account within a predetermined time period, such as 10%, 65%, etc.), fan situation C16 (i.e., the number of fans of the advertiser's official account within a predetermined time period, such as 200, 650, etc.), quality score C17 (i.e., the average score of the articles published by the advertiser's official account within a predetermined time period, such as 65, 80, etc.);

[0085] Regarding the service provider (the commodity service provider for the advertiser): Abnormal binding and unbinding C18 (i.e., the total number of times the advertiser binds or unbinds the server within a predetermined time period, such as 11 times, 20 times, etc.), abnormal recharge C19 (i.e., the total number of times the advertiser makes abnormal recharges such as large - amount consumption within a predetermined time period, such as 11 times, 20 times, etc.), service - provider account risk C20 (i.e., the total number of times the advertiser's corresponding service provider is complained of having business risks within a predetermined time period, such as 1 time, 0 times, etc.);

[0086] The associated risks include the associated risk C21 of the same entity related to risks (enterprises or groups register as one entity with the advertising operator, and then register multiple advertisers under this entity for easy management of advertisers), that is, the proportion of bad advertisers among all advertisers under the corresponding entity of the advertiser within the preset duration, such as 10%, 69%, etc.), the associated risk C22 of contact information (multiple advertisers registered with the same contact information, such as mobile phone numbers), that is, the proportion of bad advertisers among all advertisers registered with the same contact information as the advertiser within the preset duration, such as 10%, 69%, etc.).

[0087] The preset duration in the above text is the time parameter in the embodiment of the present application. Each data type corresponds to at least two initial time parameters. The initial time parameters can be T1 within 3 days, T2 within 7 days, T3 within 15 days, etc. Then, the optimal time parameter corresponding to each data type is determined according to the training result to reduce the data acquisition difficulty on the basis of ensuring the accuracy of the evaluation result; the number of initial time parameters corresponding to each data type can be the same or different, and the final time parameters corresponding to each data type can also be different.

[0088] In this embodiment, it is assumed that the initial data type Ci belongs to (C1, ……, C22, ……). For each initial data type Ci, there are multiple initial time parameters Ti (T1, ……, T3, ……). For each data type Ci and time parameter Ti, the training data S(Ui - Ci - Ti) corresponding to all sample objects Ui (including bad advertisers and good advertisers) can be obtained based on the user identification of the sample corresponding to Ui.

[0089] Then, the first screening of data parameters is carried out based on the data missing rate, the proportion of single-value data, etc. For example, for all initial time parameters under a certain data type or a certain initial time parameter, such as all initial time parameters of the Fengye complaint related to complaint information, among 20,000 sample advertisers, there are less than 10,000 advertisers with training data S(Ui-Ci-Ti) corresponding to this data type. The data missing rate of all initial time parameters under this data type is greater than 50%. The data type of the Fengye complaint related to complaint information is deleted from the initial data types. In some other scenarios, only some initial time parameters under a certain data type can be deleted. Another example is that for all initial time parameters under a certain data type or a certain initial time parameter, such as the initial time parameter (activated within 3 days) of the activation time related to account attributes, among 20,000 sample advertisers, there are less than 1,000 advertisers with training data S(Ui-Ci-Ti) corresponding to this data type being 1 (indicating that the advertiser is activated within 3 days), and the training data S(Ui-Ci-Ti) corresponding to this data type of the sample advertisers is 0 (indicating that the advertiser is not activated within 3 days). The proportion of single-value data of the initial time parameter (activated within 3 days) under this data type is greater than 95%. The initial time parameter (activated within 3 days) of the activation time related to account attributes is deleted. After the first screening, only one time parameter is retained for the remaining data types.

[0090] In this application, binning can be sub-intervals divided according to the specific values of each data type. If the value of a certain data type of an object (such as an advertiser) falls within the range of a certain sub-interval, the object can be placed in the bin represented by this sub-interval. For each data type, multiple different bin number values can be set. The initial binning parameters can be a single hot value or a value range, so that the binning of sample advertisers is as equi-depth as possible (that is, the number of sample advertisers in each bin is approximately the same). This application uses the binning method to process data, reducing the data processing difficulty. For the training data S(Ui-Ci-Ti) corresponding to a certain time parameter under a certain data type, the 20,000 values are binned respectively using multiple initial binning parameters corresponding to this time parameter under this data type, and then the binning results corresponding to each initial binning parameter (whether the number of sample advertisers in each bin is approximately equal) are checked to determine the data distribution information corresponding to each initial binning parameter, and the binning parameters with obviously uneven binning are deleted. In this way, only one binning parameter is retained for each time parameter under each data type.

[0091] After that, the data parameters are secondarily screened according to the information quantity parameter; for example, the data type is screened according to the basic variable and the information quantity parameter. For example, since abnormal login often represents the transfer of the advertiser's account, information such as the corresponding name of the advertiser will change (i.e., abnormal change), that is, the abnormal change related to the operation behavior and the abnormal login correspond to the same basic variable (abnormal operation of the advertiser). For these two data types, based on the binning result, only one data type with a larger information value (IV) of the two data types needs to be retained; another example is that there is a high correlation between the abnormal recharge related to the service provider and the risk of the service provider's account, that is, the correlation coefficient r of these two data types is greater than 0.8. At this time, based on the binning result, only one data type with a larger information value (IV) of the two data types needs to be retained.

[0092] For how to obtain the information value (IV) of a certain data type Ci, the present application provides the following method:

[0093] According to the binning parameters corresponding to the data type Ci, the data corresponding to all sample advertisers is divided into N bins, and the weight of evidence (WOE) of the i-th bin among these N bins is calculated:

[0094] ;

[0095] Among them, can be the ratio of the number of bad advertisers in the i-th bin to the number of all bad advertisers among all sample advertisers ; can be the ratio of the number of good advertisers in the i-th bin to the number of all good advertisers among all sample advertisers ;

[0096] Then, the information value (IV) of this data type is:

[0097] ;

[0098] According to the above information value (IV) and the weight of evidence (WOE), the information corresponding to each data type can be obtained, and then the data parameters can be secondarily screened according to the information quantity parameter.

[0099] In this application, the correlation coefficient r can be the Pearson product - moment correlation coefficient (commonly denoted by r), which is used to measure the degree of correlation (linear correlation) between two variables (i.e., the data types mentioned above). Its value ranges from - 1 to 1. The Pearson coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables. The specific implementation method can refer to the current method. Using the correlation coefficient r in this application can quickly determine the degree of correlation between two data types, and then screen the initial data types, minimizing the number of data types to the greatest extent and reducing the model complexity and calculation difficulty.

[0100] In this application, the classification network in the neural network can adopt the following regression model network:

[0101] ;

[0102] Among them, represents the probability that a sample advertiser is a bad advertiser. The larger the , the greater the probability of being bad. x refers to each variable entering the model. is the coefficient of this variable. The value of x is the WOE of the value of a certain data type of this advertiser belonging to the bin under the corresponding binning parameters of this data type. The training purpose of the regression model network is to make the prediction probability of the model for good advertisers among the sample advertisers less than 1 - , and the prediction probability of the model for bad advertisers among the sample advertisers greater than 1 -

[0103] In this application, the conversion network in the neural network can adopt the following conversion model network:

[0104] ;

[0105] ;

[0106] Among them, A and B are constants. The negative sign can make the lower the probability that the advertiser is a good advertiser, the higher the score, that is, a higher score value represents a higher risk that the advertiser sends advertisements with negative information.

[0107] In one embodiment, the values of the constants A and B in the formula can be obtained by substituting two known or assumed score values for calculation. Usually, two assumptions need to be set:

[0108] (1) Set a specific expected score value for a specific ratio;

[0109] (2)Determine the score for doubling the ratio (PDO)

[0110] Based on the above analysis, this application first assumes that the score for a specific point with a ratio of x is P. Then the score for the point with a ratio of 2x should be P - PDO. Substituting into the formula, the following two equations can be obtained:

[0111] ;

[0112] ;

[0113] In this embodiment, it is expected that when x = ( / 1 - ) = 5%, the score is 50 points, and PDO is 10 points (that is, for every 10 - point increase / 1 - the proportion will be reduced by half). Substituting into the formula, B = 14.43 and A = 6.78 are obtained. At this time / 1 - = 10%, Score = 40, meeting the requirements, that is:

[0114] ;

[0115] After the parameters A and B are determined, the default probability and the corresponding score of other advertisers can be predicted. For a certain advertiser, its score is:

[0116] ;

[0117] Among them, n represents the number of finally determined data types, is the WOE corresponding to the bin to which the value of the nth data type of this advertiser belongs, is positively correlated with the information amount IV corresponding to the nth data type.

[0118] Through the above steps, the training of data parameters, binning parameters, and neural networks is completed.

[0119] Steps 304 to 305: The evaluation server evaluates all advertisers and synchronizes the evaluation results to the promotion server.

[0120] In this embodiment, it is assumed that the data types after training only include five data types: penalty history C2, abnormal change C5, preliminary examination rejection C7, inspection rejection C8, and penalty history C14. The time parameter includes within 15 days. Then, the evaluation server respectively obtains the data of these types within 15 days according to the user identification of the advertiser as historical evaluation data. According to the trained binning parameters, the historical evaluation data is binned to obtain the evaluation data binning result of the promotion initiation object, that is, the WOE values corresponding to the bins to which the values of these 5 data types of each advertiser belong are obtained. The evaluation data binning result of the promotion initiation object is processed by the trained neural network to obtain the evaluation result of the promotion initiation object, that is, the prediction probability p that each advertiser is a bad advertiser is obtained by substituting the 5 WOE values of each advertiser into the regression network respectively, and further the score Score of each advertiser is obtained based on the conversion network and synchronized to the promotion server.

[0121] Steps 306 to 310: The promotion server manages the promotion information such as advertisements initiated by the advertiser according to the score of the advertiser.

[0122] In this embodiment, when an advertiser needs to initiate an advertisement, the promotion terminal is used to generate promotion information such as an advertisement and send it to the promotion server. The promotion server determines how to review this advertisement according to the score corresponding to this advertiser. For example, if the score of this advertiser is greater than 80 points, it means that the advertiser has a strong willingness to send advertisements with negative information. Then, this advertisement is sent to the review terminal for manual review. If the manual review passes, this advertisement is published. If the manual review fails, this advertisement is returned to the promotion terminal for modification, ensuring the quality of the promotion information such as advertisements. For example, if the score of this advertiser is less than 20 points, it means that the advertiser has a weak willingness to send advertisements with negative information. Then, this advertisement is sent to the review server for machine review. If the machine review passes, this advertisement is published. If the machine review fails, this advertisement is returned to the promotion terminal for modification, ensuring the review efficiency of the promotion information such as advertisements.

[0123] This embodiment completes the allocation of reasonable review resources for the advertisements initiated by the advertiser based on the evaluation result of the advertiser, ensuring the advertisement quality and review efficiency.

[0124] Figure 4 It is the third process schematic diagram of the promotion information management method provided by the embodiment of the present application. Please refer to Figure 4 This promotion information management method includes the following steps:

[0125] Steps 401 to 403: Perform the training of data parameters, binning parameters, and neural networks.

[0126] For the specific implementation of this step, please refer to the descriptions in steps 301 to 303, which completes the training of data parameters, binning parameters, and neural networks.

[0127] Steps 404 to 409: The evaluation server evaluates all advertisers and synchronizes the evaluation results to the promotion server.

[0128] In this embodiment, it is assumed that the data types after training only include: penalty history C2, abnormal change C5, preliminary review rejection C7, inspection rejection C8, penalty history C14, etc., a total of 5 data types, and the time parameter is within 15 days. Then, the evaluation server respectively obtains the data of these types within 15 days according to the user identification of the advertiser as historical evaluation data. According to the trained binning parameters, the historical evaluation data is binned to obtain the evaluation data binning result of the promotion initiation object, that is, the WOE value corresponding to the bin to which the values of these 5 data types of each advertiser belong. The evaluation data binning result of the promotion initiation object is processed by the trained neural network to obtain the evaluation result of the promotion initiation object, that is, the prediction probability p that each advertiser is a bad advertiser is obtained by substituting the 5 WOE values of each advertiser into the regression network, and further based on the conversion network, the score Score of each advertiser is obtained.

[0129] After that, set an alarm threshold, for example, alarm when the score is 50. If the score of a certain advertiser is 53, then the object information such as the score and main reasons of this advertiser is added to the alarm information and sent to the audit terminal.

[0130] After the audit terminal receives the alarm information, it is displayed in the interface as shown in Figure 8a It can be seen from Figure 8a that the alarm information includes the advertiser name, unique identifiers such as UID, industry, registration time, score, risk dimension, etc.

[0131] After the back-end personnel click on the alarm information, they enter the score display and risk reason display interface as shown in Figure 8b If the back-end personnel need to give feedback, click the feedback button in the interface as shown in Figure 8b and enter the feedback interface as shown in Figure 8c In this interface, the back-end personnel can modify some inaccurate risk reasons.

[0132] If the back-end personnel modify the score of an advertiser, the audit terminal sends the adjusted evaluation result to the evaluation server, and the evaluation server synchronously updates the evaluation result.

[0133] Steps 410 to 411: The back-end personnel obtain and display the score distribution of the advertisers.

[0134] In this embodiment, if the back-end personnel need to view the rating distribution of active advertisers in a recent period, for example, in November, a request is sent to the evaluation server through the review terminal to request the rating distribution information of the advertisers.

[0135] After that, the evaluation server sends the evaluation result to the review terminal, and the review terminal displays it in the Figure 8d shown interface. As Figure 8d shown, as the rating increases, the number of accounts decreases, indicating that most advertisers still abide by the rules. As the rating increases, the proportion of bad advertisers increases, indicating that the higher the rating of the advertiser, the more likely it is to launch ads containing negative information.

[0136] In one embodiment, the review terminal can also calculate the proportion of the scores of the overall market advertisers, and divide the scores into risk levels according to the business control threshold: high-risk interval, medium-risk interval, and low-risk interval, so as to facilitate the back-end personnel to directly understand the rating distribution of the advertisers.

[0137] In this embodiment, the risk of advertisers is measured by ratings, and advertisers with a sudden increase in ratings are alerted in a timely manner, and verified and disposed of immediately. The ratings will be updated regularly, continuously detecting the ratings of each advertiser, automatically discovering advertisers with abnormal growth rates through algorithms, the system automatically pushes abnormal alerts in the duty group, and the duty personnel verify and dispose of them in a timely manner. Based on this alert information, this embodiment can more reasonably and effectively configure the inspection logic of the ads in progress. The human and machine resources for inspecting the ads in progress are limited. In order to invest the limited resources in the most risky ad inspections and more effectively reduce the overall platform risk of the ad operators, this application uses the ratings of the advertisers as a dimension for resource allocation, increasing the inspection frequency, manpower, dispatching logic, etc. for the ads of high-risk customers, and thus more reasonably and effectively allocating the ad review manpower. In some other embodiments, the ad review human and machine resources are also limited. The ratings of the advertisers are an effective dimension for distinguishing resource allocation. For the ads of low-risk customers, more machine capabilities are used to assist in the review to improve the review efficiency of high-quality customers and improve the customer experience. For the ads of high-risk customers, more human reviews are used to avoid risky ads going online by mistake.

[0138] Correspondingly, Figure 5 is the structural schematic diagram of the promotion information management device provided by the embodiment of the present application. Please refer to Figure 5 . The promotion information management device includes the following modules:

[0139] A parameter training module 501, configured to train data parameters and binning parameters according to the training data corresponding to multiple initial data parameters of the sample object;

[0140] The data acquisition module 502 is configured to obtain historical evaluation data related to the promotion initiating object according to the data parameters;

[0141] The object evaluation module 503 is configured to bin the historical evaluation data according to the binning parameters to obtain the binned result of the evaluation data of the promotion initiating object, and obtain the evaluation result of the promotion initiating object according to the binned result of the evaluation data of the promotion initiating object;

[0142] The resource allocation module 504 is configured to allocate corresponding promotion information review resources to the promotion initiating object according to the evaluation result;

[0143] The information management module 505 is configured to review and manage the promotion information corresponding to the promotion initiating object by using the promotion information review resources.

[0144] In one embodiment, the parameter training module 501 is specifically configured to: obtain a plurality of initial data parameters and a plurality of initial binning parameters; obtain the evaluation data corresponding to the plurality of initial data parameters of the sample user and determine it as training data; train the plurality of initial data parameters and the plurality of initial binning parameters according to the training data to obtain the trained data parameters and the trained binning parameters; the trained data parameters belong to the plurality of initial data parameters.

[0145] In one embodiment, the parameter training module 501 is specifically configured to: obtain a plurality of initial data types, a plurality of initial time parameters corresponding to each initial data type, and a plurality of initial binning parameters corresponding to each initial data type; obtain the first identifier of the first sample object corresponding to the first evaluation result and the second identifier of the second sample object corresponding to the second evaluation result, where the first evaluation result and the second evaluation result are different; obtain the training data according to the first identifier, the second identifier, the plurality of initial data types, and the plurality of initial time parameters corresponding to each initial data type; perform a first data parameter screening on the plurality of initial data types and the plurality of initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; perform a first training data screening on the training data according to the screening result of the first data parameter screening; perform a binning process on the screening result of the first training data screening according to the plurality of initial binning parameters corresponding to each initial data type; perform a binning parameter screening on the plurality of initial binning parameters according to the binning result corresponding to each initial binning parameter to obtain the trained binning parameters; obtain the index parameters of the binning result of the trained binning parameters; perform a second data parameter screening on the plurality of initial data types and the plurality of initial time parameters corresponding to each initial data type according to the index parameters to obtain the trained data parameters.

[0146] In one embodiment, the parameter training module 501 is specifically configured to: determine the data missing rate of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; perform a first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the data missing rate of each initial time parameter under each initial data type.

[0147] In one embodiment, the parameter training module 501 is specifically configured to: determine the proportion of single values of the data of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; perform a first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the proportion of single values of the data of each initial time parameter under each initial data type.

[0148] In one embodiment, the parameter training module 501 is specifically configured to: determine the data distribution information of each initial time parameter of each initial data type corresponding to each initial binning parameter according to the binning result corresponding to each initial binning parameter; perform a binning parameter screening on the multiple initial binning parameters corresponding to each initial data type according to the data distribution information.

[0149] In one embodiment, the parameter training module 501 is specifically configured to: obtain the basic variables corresponding to each initial data type; obtain the information amount parameter of the training data corresponding to at least two different initial data types and the multiple initial time parameters corresponding to each initial data type for which the basic variables meet the preset conditions; perform a second data parameter screening on the at least two different initial data types that meet the preset conditions and the multiple initial time parameters corresponding to each initial data type according to the information amount parameter to obtain the trained data parameters.

[0150] In one embodiment, the parameter training module 501 is specifically configured to: obtain the information amount parameter of the training data corresponding to each initial data type and the multiple initial time parameters corresponding to each initial data type; obtain the data type correlation coefficient between each initial data type; perform a second data parameter screening on each initial data type and the multiple initial time parameters corresponding to each initial data type according to the information amount parameter to obtain the trained data parameters.

[0151] In one embodiment, the parameter training module 501 is specifically configured to: process the training data according to the trained data parameters and the trained binning parameters to obtain effective training data; use the effective training data to train the classification network to obtain a trained classification network; train the conversion network according to the display parameters corresponding to the evaluation result to obtain a trained conversion network; and determine the trained neural network according to the trained classification network and the trained conversion network.

[0152] In one embodiment, the resource allocation module 504 is specifically configured to: determine the promotion information review method corresponding to the promotion initiation object according to the association relationship between the evaluation result and the review mode and the evaluation result of the promotion initiation object; and allocate review resources using the promotion information review method corresponding to the promotion initiation object.

[0153] In one embodiment, the resource allocation module 504 is specifically configured to: determine the promotion information review resources corresponding to the promotion initiation object according to the corresponding relationship between the evaluation result and the proportion of review resources and the evaluation result of the promotion initiation object.

[0154] In one embodiment, Figure 5 The illustrated embodiment further includes a result management module 506, which is configured to: obtain an alarm threshold after the step of obtaining the evaluation result of the promotion initiation object; determine a target promotion initiation object that needs to be alarmed according to the evaluation result of the promotion initiation object and the alarm threshold; and send an alarm message to the management device, where the alarm message carries the object information of the target promotion initiation object.

[0155] In one embodiment, the result management module 506 is further configured to: receive an evaluation result query request sent by the management device, where the evaluation result query request carries the object identifier of the promotion initiation object; obtain the evaluation result of the promotion initiation object corresponding to the object identifier and the evaluation result related description information according to the object identifier; and send the evaluation result of the promotion initiation object and the evaluation result related description information to the management device, so that the management device displays the evaluation result of the promotion initiation object and the evaluation result related description information on the information display interface.

[0156] In one embodiment, the result management module 506 is further configured to: receive the feedback information sent by the management device; correct the evaluation result of the promotion initiation object according to the feedback information to obtain a corrected evaluation result; and manage the promotion information corresponding to the promotion initiation object according to the corrected evaluation result.

[0157] Correspondingly, an embodiment of the present application further provides a computer device, which includes a server or a terminal, etc.

[0158] As shown Figure 6 in the figure, the computer device may include a radio frequency (RF) circuit 601, a memory 602 including one or more computer-readable storage media, an input unit 603, a display unit 604, a sensor 605, an audio circuit 606, a wireless fidelity (WiFi) module 607, a processor 608 including one or more processing cores, and a power supply 609 and other components. Those skilled in the art can understand that Figure 6 the structure of the computer device shown in

[0159] is not a limitation on the computer device, and it may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements. Among them:

[0160] The RF circuit 601 can be used for receiving and sending information or signals during a call. Specifically, after receiving the downlink information from the base station, it is handed over to one or more processors 608 for processing; in addition, data related to the uplink is sent to the base station. The memory 602 can be used to store software programs and modules, and the processor 608 executes various functional applications and data processing by running the software programs and modules stored in the memory 602. The input unit 603 can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0161] The computer device may further include at least one sensor 605, such as a light sensor, a motion sensor, and other sensors. The audio circuit 606 includes a speaker, and the microphone can provide an audio interface between the user and the computer device.

[0162] WiFi belongs to short-distance wireless transmission technology. The computer device can help users send and receive emails, browse the web, and access streaming media through the WiFi module 607, and it provides users with wireless broadband Internet access. Although Figure 6 the WiFi module 607 is shown, it can be understood that it does not belong to an essential component of the computer device and can be omitted completely within the scope of not changing the essence of the application as needed.

[0163] The processor 608 is the control center of the computer device, connecting various parts of the entire mobile phone through various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 602, and by invoking the data stored in the memory 602, it executes various functions of the computer device and processes data, thereby monitoring the mobile phone as a whole.

[0164] The computer device also includes a power supply 609 (such as a battery) for powering each component. Preferably, the power supply can be logically connected to the processor 608 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system.

[0165] Although not shown, the computer device may also include a camera, a Bluetooth module, etc., which will not be elaborated here. Specifically in this embodiment, the processor 608 in the computer device will load the executable files corresponding to the processes of one or more application programs into the memory 602 according to the following instructions, and the processor 608 will run the application programs stored in the memory 602, so as to realize the following functions:

[0166] Train to obtain data parameters and binning parameters according to the training data corresponding to multiple initial data parameters of the sample object;

[0167] Obtain historical evaluation data related to the promotion initiation object according to the data parameters;

[0168] Bin the historical evaluation data according to the binning parameters to obtain the binning result of the evaluation data of the promotion initiation object, and obtain the evaluation result of the promotion initiation object according to the binning result of the evaluation data of the promotion initiation object;

[0169] Allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result;

[0170] Use the promotion information review resources to review and manage the promotion information corresponding to the promotion initiation object.

[0171] In the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not elaborated in a certain embodiment, reference can be made to the detailed description above, which will not be elaborated here.

[0172] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling related hardware through instructions. These instructions can be stored in a computer-readable storage medium and loaded and executed by the processor.

[0173] To this end, an embodiment of the present application provides a computer-readable storage medium, which stores multiple instructions that can be loaded by a processor to implement the following functions:

[0174] Train data parameters and binning parameters based on training data corresponding to multiple initial data parameters of a sample object;

[0175] Obtain historical evaluation data related to a promotion initiation object according to the data parameters;

[0176] Bin the historical evaluation data according to the binning parameters to obtain a binned result of the evaluation data of the promotion initiation object, and obtain an evaluation result of the promotion initiation object according to the binned result of the evaluation data of the promotion initiation object;

[0177] Allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result;

[0178] Use the promotion information review resources to review and manage the promotion information corresponding to the promotion initiation object.

[0179] For the specific implementation of each of the above operations, reference may be made to the previous embodiments, which will not be elaborated herein.

[0180] Among them, the storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc.

[0181] Since the instructions stored in the storage medium can execute the steps in any method provided by the embodiments of the present application, the beneficial effects that can be achieved by any method provided by the embodiments of the present application can be realized. For details, reference may be made to the previous embodiments, which will not be elaborated herein.

[0182] At the same time, an embodiment of the present application provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above various optional implementation manners. For example, to implement the following functions:

[0183] Train data parameters and binning parameters based on training data corresponding to multiple initial data parameters of a sample object;

[0184] Obtain historical evaluation data related to a promotion initiation object according to the data parameters;

[0185] Bin the historical evaluation data according to the binning parameters to obtain the binning result of the evaluation data of the promotion initiation object, and obtain the evaluation result of the promotion initiation object according to the binning result of the evaluation data of the promotion initiation object;

[0186] Allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result;

[0187] Use the promotion information review resources to review and manage the promotion information corresponding to the promotion initiation object.

[0188] The above has introduced in detail a promotion information management method, device, computer device and computer-readable storage medium provided by an embodiment of the present application. Specific examples are used in this article to elaborate on 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 skilled 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 promotion information management method, characterized in that, Including: Training data corresponding to multiple initial data parameters of a sample object to obtain data parameters and binning parameters; Obtaining historical evaluation data related to a promotion initiation object according to the data parameters; Binning the historical evaluation data according to the binning parameters to obtain a binning result of the evaluation data of the promotion initiation object, and obtaining an evaluation result of the promotion initiation object according to the binning result of the evaluation data of the promotion initiation object; Allocating corresponding promotion information review resources for the promotion initiation object according to the evaluation result; Using the promotion information review resources to review and manage the promotion information corresponding to the promotion initiation object; The step of training data parameters and binning parameters by training data corresponding to multiple initial data parameters of a sample object includes: sequentially performing a first data parameter screening, a first training data screening, a binning process, and a binning parameter screening on the training data to obtain trained binning parameters, and then performing a second data parameter screening according to the trained binning parameters to obtain trained data parameters. The screening condition of the first data parameter screening includes a data missing rate or a proportion of single values in the data, and the screening condition of the second data parameter screening includes an information amount parameter or a data type correlation coefficient.

2. The promotion information management method according to claim 1, wherein The step of training data parameters and binning parameters by training data corresponding to multiple initial data parameters of a sample object includes: Obtaining multiple initial data parameters and multiple initial binning parameters; Obtaining evaluation data corresponding to multiple initial data parameters of the sample object and determining the training data corresponding to multiple initial data parameters of the sample object; Training the multiple initial data parameters and multiple initial binning parameters according to the training data to obtain trained data parameters and trained binning parameters; the trained data parameters belong to the multiple initial data parameters.

3. The promotion information management method according to claim 2, characterized in that The data parameters include a data type and a time parameter; The step of training data parameters and binning parameters by training data corresponding to multiple initial data parameters of a sample object includes: Obtaining multiple initial data types, multiple initial time parameters corresponding to each initial data type, and multiple initial binning parameters corresponding to each initial data type; Obtaining a first identifier of a first sample object corresponding to a first evaluation result and a second identifier of a second sample object corresponding to a second evaluation result, where the first evaluation result and the second evaluation result are different; Obtaining the training data according to the first identifier, the second identifier, the multiple initial data types, and the multiple initial time parameters corresponding to each initial data type; Performing a first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; Performing a first training data screening on the training data according to the screening result of the first data parameter screening; Performing a binning process on the screening result of the first training data screening according to the multiple initial binning parameters corresponding to each initial data type; Perform binning parameter screening on the multiple initial binning parameters according to the binning results corresponding to the respective initial binning parameters to obtain the trained binning parameters; Obtain the metric parameters of the binning results of the trained binning parameters; According to the metric parameters, perform second data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type to obtain the trained data parameters.

4. The promotion information management method according to claim 3, wherein The step of performing first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data includes: Determine the data missing rate of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; According to the data missing rates of each initial time parameter under each initial data type, perform first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type.

5. The promotion information management method according to claim 3, wherein The step of performing first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data includes: Determine the proportion of single values of the data of each initial time parameter under each initial data type according to the training data content corresponding to each initial time parameter under each initial data type in the training data; According to the proportion of single values of the data of each initial time parameter under each initial data type, perform first data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type.

6. The promotion information management method according to claim 3, characterized in that The step of performing binning parameter screening on the multiple initial binning parameters according to the binning results corresponding to the respective initial binning parameters to obtain the trained binning parameters includes: Determine the data distribution information of each initial time parameter of each initial data type corresponding to each initial binning parameter according to the binning results corresponding to the respective initial binning parameters; According to the data distribution information, perform binning parameter screening on the multiple initial binning parameters corresponding to each initial data type.

7. The promotion information management method according to claim 3, characterized in that The metric parameters include information quantity parameters. The step of performing second data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the metric parameters to obtain the trained data parameters includes: Obtain the basic variables corresponding to each initial data type; Obtain the information quantity parameters of the training data corresponding to at least two different initial data types and the multiple initial time parameters corresponding to each initial data type for which the basic variables meet the preset conditions; According to the information quantity parameters, perform second data parameter screening on the at least two different initial data types that meet the preset conditions and the multiple initial time parameters corresponding to each initial data type to obtain the trained data parameters.

8. The promotion information management method according to claim 3, wherein The indicator parameters include information quantity parameters and data type correlation coefficients. The step of performing a second data parameter screening on the multiple initial data types and the multiple initial time parameters corresponding to each initial data type according to the indicator parameters to obtain the trained data parameters includes: Obtaining the information quantity parameters of the training data corresponding to each initial data type and the multiple initial time parameters corresponding to each initial data type; Obtaining the data type correlation coefficients between each initial data type; Performing a second data parameter screening on each initial data type and the multiple initial time parameters corresponding to each initial data type according to the information quantity parameters and the data type correlation coefficients to obtain the trained data parameters.

9. The promotion information management method according to claim 1, wherein The step of obtaining the evaluation result of the promotion initiation object according to the evaluation data binning result of the promotion initiation object includes: Obtaining the trained neural network; Using the trained neural network to process the evaluation data binning result of the promotion initiation object to obtain the evaluation result of the promotion initiation object.

10. The promotion information management method according to claim 1, wherein The step of allocating corresponding promotion information review resources for the promotion initiation object according to the evaluation result includes: Determining the promotion information review method corresponding to the promotion initiation object according to the association relationship between the evaluation result and the review mode and the evaluation result of the promotion initiation object; Using the promotion information review method corresponding to the promotion initiation object to allocate corresponding promotion information review resources for the promotion initiation object.

11. The promotion information management method according to claim 1, characterized in that The step of allocating corresponding promotion information review resources for the promotion initiation object according to the evaluation result includes: Determining the promotion information review resources corresponding to the promotion initiation object according to the corresponding relationship between the evaluation result and the review resource ratio and the evaluation result of the promotion initiation object.

12. The promotion information management method according to claim 1, wherein After the step of obtaining the evaluation result of the promotion initiation object, it further includes: Obtaining an alarm threshold; Determining a target promotion initiation object that needs to be alarmed according to the evaluation result of the promotion initiation object and the alarm threshold; Sending an alarm message to the management device, where the alarm message carries the object information of the target promotion initiation object.

13. The promotion information management method according to any one of claims 1 to 12, characterized in that After the step of obtaining the evaluation result of the promotion initiation object, it further includes: Receiving an evaluation result query request sent by the management device, where the evaluation result query request carries the object identifier of the promotion initiation object; Obtaining the evaluation result of the promotion initiation object corresponding to the object identifier and the evaluation result related description information according to the object identifier; Sending the evaluation result of the promotion initiation object and the evaluation result related description information to the management device so that the management device displays the evaluation result of the promotion initiation object and the evaluation result related description information on the information display interface.

14. The promotion information management method according to claim 13, characterized in that After the step of sending the evaluation result of the promotion initiation object and the evaluation result related description information to the management device, it further includes: Receiving the feedback information sent by the management device; Correcting the evaluation result of the promotion initiation object according to the feedback information to obtain the corrected evaluation result; Manage the promotion information corresponding to the promotion initiation object according to the corrected evaluation result.

15. An apparatus for managing promotion information, characterized in that, Including: A parameter training module, configured to train data parameters and binning parameters according to training data corresponding to multiple initial data parameters of a sample object; A data acquisition module, configured to acquire historical evaluation data related to the promotion initiation object according to the data parameters; An object evaluation module, configured to bin the historical evaluation data according to the binning parameters to obtain a binned result of the evaluation data of the promotion initiation object, and obtain an evaluation result of the promotion initiation object according to the binned result of the evaluation data of the promotion initiation object; A resource allocation module, configured to allocate corresponding promotion information review resources for the promotion initiation object according to the evaluation result; An information management module, configured to review and manage the promotion information corresponding to the promotion initiation object by using the promotion information review resources; The parameter training module is specifically configured to: sequentially perform a first data parameter screening, a first training data screening, a binning process, and a binning parameter screening on the training data to obtain the trained binning parameters, and then perform a second data parameter screening according to the trained binning parameters to obtain the trained data parameters. The screening condition for the first data parameter screening includes the data missing rate or the proportion of single values in the data, and the screening condition for the second data parameter screening includes the information amount parameter or the data type correlation coefficient.

16. A computer device, characterized in that, Including a memory and a processor, where a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps in the promotion information management method according to any one of claims 1 to 14.

17. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in the promotion information management method according to any one of claims 1 to 14.

18. A computer program product, characterized in that, Including computer instructions, and the computer instructions are loaded by a processor to execute the steps in the promotion information management method according to any one of claims 1 to 14.

Images