Information push method, device, electronic device and readable storage medium

Abstract

The embodiments of the present application provide an information push method, device, electronic device and computer-readable storage medium, which relate to the field of data processing. The method includes: obtaining a candidate push information set, a push strategy corresponding to the candidate push information set and a first proportion corresponding to each push strategy; determining the push ranking results of each initial candidate push information in the candidate push information set under each push strategy; randomly determining the target push strategy corresponding to the target candidate push information to be selected from the push strategy; based on the target push strategy corresponding to the target push information to be selected, the target candidate push information is selected from each initial candidate push information according to the push ranking results corresponding to the target push strategy. The embodiments of the present application ensure diversity when obtaining target candidate push information, ensure fair competition among strategies, and are conducive to objective and efficient evaluation of strategy effects.

Description

Technical Field

[0001] The present application relates to the field of data processing technology. Specifically, the present application relates to an information push method, device, electronic device and readable storage medium. Background Art

[0002] In order to improve the diversity of recommendation results, multi-strategy fusion is a typical problem faced by online search and recommendation systems. At present, the way to improve the diversity of recommendation information in the industry is mainly to establish recall branches according to different optimization goals in the recall stage.

[0003] The recall results are directly merged and pre-sorted by rough sorting before being returned to fine sorting. The rough sorting process is responsible for efficient and lightweight sorting and truncation of the recall information, screening out a small amount of relatively high-quality information (in the hundreds) for fine sorting. Given the limited length of the fine sorting queue, how to ensure that different recall branch ads have the opportunity to enter the fine sorting and compete for exposure opportunities is a challenge in the rough sorting process. Summary of the invention

[0004] The embodiments of the present application provide an information push method, device, electronic device, computer-readable storage medium, and computer program product, which can solve the above-mentioned problems of the prior art. The technical solution is as follows:

[0005] According to one aspect of an embodiment of the present application, there is provided an information push method, the method comprising:

[0006] and a first proportion corresponding to each push strategy, the candidate push information set includes at least one initial candidate push information; the first proportion is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to the preset number;

[0007] Determine the push ranking results of each initial candidate push information in the candidate push information set under each push strategy;

[0008] According to the first proportion, randomly determine a target push strategy corresponding to each target candidate push information to be selected from at least one push strategy;

[0009] Based on the target push strategy corresponding to each target push information to be selected, and according to the push ranking result corresponding to the target push strategy, select the target candidate push information from each initial candidate push information;

[0010] Determine the target push information from the target candidate push information, and push the target push information to the target object

[0011] As an optional embodiment, according to the first proportion, randomly determining a target push strategy corresponding to each target candidate push information to be selected from at least one push strategy includes:

[0012] Determine the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy;

[0013] Initialize a target candidate push information sequence, where each element in the target candidate push information sequence represents a target candidate push information to be selected;

[0014] According to the number of target candidate push information corresponding to each push strategy, a corresponding number of elements are randomly selected from the target candidate push information sequence, and the corresponding push strategies are marked. The marked push strategies represent the target push strategies corresponding to the corresponding elements.

[0015] As an optional embodiment, selecting target candidate push information from each initial candidate push information according to the push ranking result corresponding to the target push strategy includes:

[0016] Determine reference information, where the reference information is the selected target candidate push information;

[0017] According to the push ranking result corresponding to the target push strategy, the target candidate push information corresponding to the current operation is selected from the information in each candidate push information except the reference information.

[0018] As an optional embodiment, obtaining at least one push strategy corresponding to the candidate push information set and a first proportion corresponding to each push strategy includes:

[0019] Obtaining diversified configuration information, the diversified configuration information including at least one policy group and a second proportion of each policy group, each policy group including at least one push policy, the second proportion representing the proportion of the number of target candidate push information corresponding to the corresponding policy group to a preset number;

[0020] Determine a third proportion corresponding to each push policy in each policy group, where the third proportion is used to represent the proportion of the target candidate push information selected by the corresponding push policy to the target candidate push information selected by all push policies in the policy group to which the corresponding push policy belongs;

[0021] For each push policy, the first proportion of the push policy is determined according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group.

[0022] As an optional embodiment, determining the third proportion of policies in each policy group includes:

[0023] Obtain object information of the target object of the information to be pushed;

[0024] According to the object information, a third proportion corresponding to the object information is determined.

[0025] As an optional embodiment, the object information includes feedback information for historical push information;

[0026] Determining the third proportion corresponding to the object information includes:

[0027] Determining a third proportion corresponding to the feedback information of the target object from the predetermined proportion configuration information;

[0028] The proportion configuration information includes a proportion relationship corresponding to at least one feedback information in a one-to-one manner; the proportion relationship is used to characterize the third proportion of each strategy in the strategy group.

[0029] As an optional embodiment, obtaining the diversified configuration information also includes:

[0030] Create at least one policy group and initialize the third percentage of each push policy in the policy group;

[0031] In response to an update of the third proportion of at least one push policy in the policy group or an update of the number of push policies in the policy group, determining the third proportions of all push policies in the updated policy group;

[0032] If the sum of the third proportions of all push policies in the updated policy group is less than the preset value, then increase the third proportion of at least one push policy in the policy group until the third proportions of all push policies in the policy group are equal to the preset value;

[0033] If the sum of the third proportions of all push policies in the updated policy group is greater than the preset value, the updated third proportion of at least one push policy in the policy group is compressed until the third proportions of all push policies in the policy group are equal to the preset value.

[0034] According to another aspect of an embodiment of the present application, there is provided an information push device, the device comprising:

[0035] An information acquisition module is used to acquire a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, wherein the candidate push information set includes at least one initial candidate push information; the first proportion is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to a preset number;

[0036] A single strategy sorting module determines the push sorting results of each initial candidate push information in the candidate push information set under each push strategy;

[0037] A strategy shuffling module, configured to randomly determine a target push strategy corresponding to each target candidate push information to be selected from at least one push strategy according to the first proportion;

[0038] A push information selection module, for selecting target candidate push information from each initial candidate push information based on the target push strategy corresponding to each target push information to be selected and according to the push sorting result corresponding to the target push strategy;

[0039] The push module is used to determine the target push information from the target candidate push information and push the target push information to the target object.

[0040] As an optional embodiment, the policy shuffling module includes:

[0041] An information quantity determination submodule, used to determine the quantity of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy;

[0042] A sequence initialization submodule is used to initialize a target candidate push information sequence, wherein each element in the target candidate push information sequence represents a target candidate push information to be selected;

[0043] The random marking submodule is used to randomly select a corresponding number of elements in the target candidate push information sequence according to the number of target candidate push information corresponding to each push strategy, and mark the corresponding push strategy. The marked push strategy represents the target push strategy corresponding to the corresponding element.

[0044] As an optional embodiment, the push information selection module includes:

[0045] The reference information determination submodule is used to determine the reference information, where the reference information is the selected target candidate push information;

[0046] The sorting and filtering submodule is used to select the target candidate push information corresponding to the current operation from the information other than the reference information in each candidate push information according to the push sorting result corresponding to the target push strategy.

[0047] As an optional embodiment, the information acquisition module includes:

[0048] A configuration information acquisition submodule is used to acquire diversified configuration information, wherein the diversified configuration information includes at least one policy group and a second proportion of each policy group, each policy group includes at least one push policy, and the second proportion represents the proportion of the number of target candidate push information corresponding to the corresponding policy group to the preset number;

[0049] A third proportion determination submodule is used to determine a third proportion corresponding to each push strategy in each strategy group, where the third proportion is used to represent the proportion of the target candidate push information selected by the corresponding push strategy to the target candidate push information selected by all push strategies in the strategy group where the corresponding push strategy belongs;

[0050] The first proportion determination submodule is used to determine, for each push policy, a first proportion of the push policy according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group.

[0051] As an optional embodiment, the object information includes feedback information for historical push information;

[0052] The third proportion determination submodule is specifically used to: determine a third proportion corresponding to the feedback information of the target object from the predetermined proportion configuration information;

[0053] The proportion configuration information includes a proportion relationship corresponding to at least one feedback information in a one-to-one manner; the proportion relationship is used to characterize the third proportion of each strategy in the strategy group.

[0054] As an optional embodiment, the push device further includes a configuration update module, and the configuration update module includes:

[0055] A policy group creation submodule is used to create at least one policy group and initialize the third proportion of each push policy in the policy group;

[0056] A policy update submodule, configured to determine the third proportions of all push policies in the updated policy group in response to an update of the third proportion of at least one push policy in the policy group or an update of the number of push policies in the policy group;

[0057] A first proportion adjustment submodule is used to increase the third proportion of at least one push policy in the policy group if the sum of the third proportions of all push policies in the updated policy group is less than a preset value, until the third proportions of all push policies in the policy group are equal to the preset value;

[0058] The second proportion adjustment submodule compresses the updated third proportion of at least one push policy in the policy group until the third proportions of all push policies in the policy group are equal to the preset value if the sum of the third proportions of all push policies in the policy group after the update is greater than the preset value.

[0059] According to another aspect of an embodiment of the present application, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned information push method.

[0060] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned information push method are implemented.

[0061] According to one aspect of an embodiment of the present application, a computer program product is provided, including a computer program, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned information push method are implemented.

[0062] The beneficial effects of the technical solution provided by the embodiment of the present application are:

[0063] By obtaining a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, the candidate push information set includes at least one initial candidate push information, and the push ranking results of each initial candidate push information in the candidate push information set under each push strategy are determined. According to the first proportion, a target push strategy corresponding to each target candidate push information to be selected is randomly determined from at least one push strategy. On the one hand, diversity is ensured when obtaining target candidate push information, and on the other hand, the problem of lack of fairness in strategy priority is solved, and fair competition among strategies is ensured, which is conducive to objective and efficient evaluation of strategy effects. It not only saves time for troubleshooting strategy coverage problems when strategy priorities are repeatedly aligned and strategy effects do not meet expectations, but also provides a unified effect evaluation mechanism at the framework level for long-term assessment of strategy effects, thereby further improving the optimization efficiency of the push system. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0065] Figure 1 A schematic diagram of a process for determining target candidate push information for the prior art;

[0066] Figure 2 A flowchart of the core concept of the information push method provided in the embodiment of the present application;

[0067] Figure 3 A schematic diagram of a system architecture for implementing information push provided in an embodiment of the present application;

[0068] Figure 4 A flowchart of an information push method provided in an embodiment of the present application;

[0069] Figure 5 A schematic diagram of a process for randomly determining a target push strategy corresponding to each target candidate push information to be selected provided in an embodiment of the present application;

[0070] Figure 6 A flowchart of an information push method provided in yet another embodiment of the present application;

[0071] Figure 7A schematic diagram of diversified configuration information provided by an embodiment of the present application;

[0072] Figure 8 A flowchart of an information push method provided in another embodiment of the present application;

[0073] Fig. 9 This is a flowchart of an information push method according to another embodiment of the present application;

[0074] Fig.10 A schematic diagram of the structure of an information push device provided in an embodiment of the present application;

[0075] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0076] The embodiments of the present application are described below in conjunction with the drawings in the present application. It should be understood that the implementation methods described below in conjunction with the drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions of the embodiments of the present application.

[0077] It will be understood by those skilled in the art that, unless specifically stated, the singular forms "one", "an" and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application refer to that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude the implementation as other features, information, data, steps, operations, elements, components and / or combinations thereof supported by the technical field. It should be understood that when we call an element "connected" or "coupled" to another element, the one element can be directly connected or coupled to the other element, or it can refer to that the one element and the other element establish a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used here may include wireless connection or wireless coupling. The term "and / or" used here indicates at least one of the items defined by the term, for example, "A and / or B" can be implemented as "A", or as "B", or as "A and B".

[0078] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0079] First, several terms involved in this application are introduced and explained:

[0080] Rough sorting: Limited by computing resources and response latency (hundreds of milliseconds), in order to balance computing performance and push effects, push systems usually adopt a phased funnel-shaped system architecture, including recall -> rough sorting -> fine sorting. Rough sorting is responsible for efficient and lightweight sorting and truncation of the retrieval recall advertisement queue, screening out a small amount of relatively high-quality target candidate push information (in the hundreds) for fine sorting, and at the same time, it must also have the ability to explore and discover potential high-quality information.

[0081] Push strategy: Given a set of ads, a scheme for ranking the ads, such as sorting by eCPM, sorting by interest relevance score, etc. For simplicity, it is referred to as "strategy" in this article. A push strategy solves two problems: which ads to score and how to score the ads.

[0082] Diversity sorting: Given a set of initial candidate push information, the initial candidate push information is sorted and truncated under the joint action of multiple push strategies.

[0083] Diversity sorting strategy framework: A functional module in the sorting system that supports diversity sorting. It provides sorting strategies and interface definitions for the relationships between strategies. In the diversity sorting strategy framework, diversity sorting is supported in a bucketing manner. Each push strategy corresponds to a bucket. Information that meets the strategy requirements is placed in the corresponding bucket (one information may belong to multiple buckets). The information in the bucket is sorted according to the algorithm specified by the strategy to form the information queue of the strategy. Finally, the top advertisements are selected from the queues of multiple strategies and added to the final result queue. The number of information selected in each strategy queue is defined by the strategy quota.

[0084] Bucket: A collection of ads with certain characteristics. Ads in the same bucket are usually processed by the same sorting strategy.

[0085] Quota: The maximum number of sorting queue positions allowed for a certain policy, expressed as a percentage of the total queue length.

[0086] eCPM (effective cost per mille) refers to the revenue that can be obtained for every thousand impressions, and the unit of impression can be a web page.

[0087] eCPM strategy: A ranking strategy that aims to measure eCPM. The core focus of this type of strategy is to obtain a more accurate eCPM estimate or ranking relationship for an ad. Typical examples include: the existing strategy of calculating eCPM through lite / pCTR*lite / pCVR*bid, and the ranking strategy that uses the model estimation results of the fitted precise ranking eCPM. eCPM strategy is the mainstream ranking strategy for most traffic.

[0088] Click-Through Rate (pCTR) prediction is to predict the click probability of a specific object in a specific push information position under a specific environment.

[0089] Conversion rate prediction (PredictClick Value Rate, pCVR), in short, is the conversion rate from an object clicking on a push message to becoming an effectively activated, registered, or even paying object.

[0090] Non-eCPM strategies: ranking strategies that do not directly use eCPM as the optimization goal. For example, the contract guarantee and precision ranking strategy is a strategy set up to meet strong business operation requirements, and the interest-breaking exploration strategy is a strategy set up to broaden the target's vision and explore the target's interests. For the push system, eCPM maximization is the core goal of system optimization, and the existence of non-eCPM strategies is often for the long-term benefits of push information ecosystem construction and target experience.

[0091] In order to improve the diversity of recommendation results, multi-strategy fusion is a typical problem faced by online search and recommendation systems. At present, the way for the industry to improve the diversity of push information is mainly to establish recall branches according to different optimization goals in the recall stage.

[0092] The recall results are directly merged and pre-sorted by rough sorting before being returned to fine sorting. The rough sorting process is responsible for efficient and lightweight sorting and truncation of the recalled push information, screening out a small amount of relatively high-quality push information (in the hundreds) for fine sorting. Given the limited length of the fine sorting queue, how to ensure that different recall branch advertisements have the opportunity to enter the fine sorting and compete for exposure opportunities is a challenge in the rough sorting process. Existing technologies include:

[0093] 1) Score ads according to different push strategies, and merge the scores into one score (weighted average, ranking fusion, etc.). Different push strategies are reflected as different scoring factors. This method can merge different scoring bases into a single factor, and the system is simple to implement, but it has the following shortcomings:

[0094] A. Push strategies have different dimensions and are difficult to merge. Even if they are merged after simple maximum and minimum value normalization, there may still be problems with insufficient interpretation of the validity of the physical meaning after fusion.

[0095] B. Difficult to attribute. The weight can reflect the impact of the push strategy on the scoring results to a certain extent, but a large impact (large weight) may not necessarily describe the impact of the push strategy on the system. For example, when a new push strategy is added, the new push strategy is added to the scoring formula as an independent scoring factor. The weight of the new scoring factor is not the largest, but the addition of this independent factor is enough to improve the ranking of the new push information and increase the number of high-potential new push information entering the refined ranking. It can be seen that it is not comprehensive to judge the impact of the push strategy based solely on the weight of the scoring factor.

[0096] 2) Different push strategies are sorted independently, and the push strategy quota and merge order are specified, and the results are merged in the specified order.

[0097] See also Figure 1 , which exemplarily shows a flow chart of determining target candidate push information in the prior art. As shown in the figure, for the initial candidate push information Ad1 to Adn, if there are two push strategies (push strategy 1 and push strategy 2) in the current scenario, the push ranking results of the initial candidate push information under the two push strategies are obtained respectively. The quotas of push strategy 1 and push strategy 2 are specified to be 60% and 40% respectively, and it is determined that the priority of push strategy 1 is higher than the priority of push strategy 2. Then, when 10 target candidate push information are finally obtained, the first 6 target candidate push information are obtained from the push ranking result of push strategy 1, and the last 4 target candidate push information are obtained from the push ranking result of push strategy 2. If it is further determined that the target candidate push information selected by different push strategies is not repeated, then when selecting the target candidate push information from the push ranking result of push strategy 2, it is necessary to avoid the target candidate push information selected based on push strategy 1.

[0098] Depend on Figure 1 It can be seen that although this method controls the impact of different push strategies through the quota method, the sequential merging method also introduces the problem of push strategy priority: on the one hand, high-priority strategies may select high-quality information first, causing low-priority strategies to be at a disadvantage in the posterior analysis of the effect. However, in actual application, the effect is not the only criterion for judging the pros and cons of push strategies. For two push strategies with similar effects, the system should choose the one with higher performance. On the other hand, this method will also be coupled with the strategies of the dependent order relationship in the system, which will eventually lead to inconsistency between the online effective quota and the configuration. As shown in the following table, the Lite and LTR quotas are configured to be 3:1, but the final effective quotas are different under different priorities, and are inconsistent with the configuration.

[0099] The information push method, device, electronic device, computer-readable storage medium and computer program product provided in this application are intended to solve the above technical problems in the prior art.

[0100] The following describes several exemplary embodiments to illustrate the technical solutions of the embodiments of the present application and the technical effects produced by the technical solutions of the present application. It should be noted that the following embodiments can refer to, draw on or combine with each other, and the same terms, similar features and similar implementation steps in different embodiments will not be described repeatedly.

[0101] Figure 2 This is a flow chart of the core concept of the information push method of an embodiment of the present application. As shown in the figure, a candidate push information set (including n initial candidate push information), m push strategies corresponding to the candidate push information set, and a first proportion corresponding to each push strategy are obtained. The first proportion is the ratio of the number of target candidate push information to be selected corresponding to the corresponding push strategy to a preset number.

[0102] Determine each initial candidate push information in the candidate push information set, the push ranking result under each push strategy, and randomly cross-sort the push ranking results under each push strategy according to the first proportion to obtain a preset number of target candidate push information.

[0103] The push sorting results under each push strategy are randomly cross-mixed according to the first proportion, which means that on the one hand, the target candidate push information determined based on each push strategy accounts for a preset number of proportions corresponding to the corresponding push strategy, and on the other hand, the order of the target candidate push information determined based on different push strategies is random. For example, the first target candidate push information is determined to be based on strategy one, the second target candidate push information is based on strategy four, the third target candidate push information is based on strategy two, and so on, which is the above-mentioned "random cross-mixing".

[0104] The embodiments of the present application can circumvent the problems caused by the priorities in the prior art, so that the push strategies formulated by various policy parties can be verified and compared in a fairer system environment. On the one hand, it saves the time for repeatedly aligning the policy priorities and troubleshooting the sorting policy coverage issues when the policy effects do not meet expectations. On the other hand, it also provides a unified effect evaluation mechanism at the framework level for long-term assessment of the effectiveness of the strategy.

[0105] Figure 3 A schematic diagram of a system architecture for implementing information push provided in an embodiment of the present application is shown in the figure. As shown in the figure, the system mainly includes the following modules:

[0106] 1) Retrieval module: When a push request is generated, the retrieval module retrieves initial candidate push information that meets the targeting conditions from a push information library of millions. The scale of the initial candidate push information is in the tens of thousands.

[0107] 2) Coarse sorting module: responsible for performing efficient and lightweight sorting and truncation of the initial candidate push information queue recalled by the retrieval module, screening out a small number of relatively high-quality target candidate pushes (scale in the hundreds) and giving them to the fine sorting module. At the same time, it must also have the ability to explore and discover potential high-quality information.

[0108] The coarse sorting module of the embodiment of the present application includes at least one single-path sorting module and a multi-path merging module. Since the magnitude of the initial candidate push information is still large, the embodiment of the present application can set up multiple parallel single-path sorting modules to obtain the target candidate push information. For example, according to whether the last digit of the unique identifier of the initial candidate push information is an odd number or an even number, two single-path sorting modules are set to process the initial candidate push information whose last digit of the unique identifier is an odd number or an even number respectively.

[0109] By counting the feedback information of the object for historical push information, the object's historical preference for push information (such as clicks, exposure, registration, etc.) can be determined, and the offline score can be quantified, that is, the offline score can be used for representation; and using the feedback information of historical push information as a training sample, model training can also be performed to obtain an online prediction model that predicts the current or future object's preference for push information. Each single-path sorting module selects push information that meets the object's preference from the initial candidate push information by calling the offline score or online prediction model in the scoring module, and uses a diversified sorting strategy framework to obtain the target candidate push information corresponding to each single-path sorting module for the push information that meets the object's preference. The multi-path merging module sorts the target candidate push information obtained by all single-path sorting modules, and retains the top-ranked target candidate push information and inputs it into the fine sorting module.

[0110] The single-path sorting module, the multi-path merging module and the refined sorting module can save the input and / or output push information in the log to facilitate subsequent analysis.

[0111] 3) Fine ranking module: responsible for bidding and billing, sorting the small amount of high-quality target candidate push information returned by the rough ranking module according to the eCPM value, and selecting the top few high eCPM target candidate push information as the target push information. The fine ranking module counts the effect information of each strategy and saves it to the push information log for business reporting.

[0112] The diversity ranking strategy framework is mainly used in coarse ranking modules, such as Figure 3 As shown, the framework mainly includes:

[0113] 1) Policy configuration module: responsible for policy registration, preset and dynamic settings.

[0114] The policy configuration module further includes a policy registration module and a policy configuration acquisition module. The policy registration module can store the policy generator registration push policy and generate a unique identifier for the policy. By loading the policy proportion initialization configuration file, the initial third proportion of each policy can be determined. The policy configuration acquisition module can store the second proportion, the third proportion, and generate the first proportion of the policy based on the second proportion and the third proportion. The policy configuration acquisition module can initialize the policy quota based on the offline data analysis conclusion, and also supports dynamic adjustment of the policy quota.

[0115] 2) Single strategy sorting module: responsible for obtaining the sorting factor of the single strategy, sorting the initial candidate push information, and generating the sorting result corresponding to the single strategy.

[0116] 3) Strategy shuffling module: Generates random position queues according to policy quotas, shuffles and truncates multiple sorting results, and obtains the target candidate push information queue.

[0117] The present application provides an information push method, such as Figure 4 As shown, the method includes:

[0118] S101: Obtain a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy.

[0119] It should be noted that the candidate push information set of the embodiment of the present application includes at least one initial candidate push information, which can be retrieved from a million-level information library according to the object information of the target object, the context information of the currently browsed web page, etc., and the scale of these initial candidate push information is in the tens of thousands. For example, if the target object is a sports enthusiast, the initial candidate push information obtained can be information related to sports. For example, if the target object is browsing a short video of newborn feeding, the initial candidate push information obtained can be information related to parenting.

[0120] The first proportion of this application is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to the preset number. The preset number is the number of target candidate push information finally obtained. This application obtains target candidate push information based on each push strategy, and the proportion of the preset number is the first proportion corresponding to the corresponding push strategy. For example, if the first proportion of a certain push strategy is 10%, it means that the number of target candidate push information selected based on the push strategy accounts for 10% of the preset number.

[0121] The push strategies obtained in the embodiments of the present application may include ECPM push strategies (i.e., determining target candidate push information with the goal of maximizing ECPM), push strategies based on the similarity of the object's historical click information (i.e., determining target candidate push information with the goal of obtaining the maximum similarity with the object's historical click information), push strategies based on high click rates of similar objects, etc. The embodiments of the present application do not make any specific limitations on the push strategies obtained.

[0122] S102: Determine the push ranking result of each initial candidate push information in the candidate push information set under each push strategy.

[0123] The embodiment of the present application can obtain the push ranking results of each initial candidate push information under each push strategy. Generally, the higher the ranking of the initial candidate push information, the more likely it is to be selected as the target candidate push information. The order of the same initial candidate push information in the ranking results corresponding to different strategies may be the same or different.

[0124] S103: According to the first proportion, randomly determine a target push strategy corresponding to each target candidate push information to be selected from at least one push strategy.

[0125] It should be understood that if all the acquired target candidate push information constitute a sequence, for the method of setting different priorities for strategies in the prior art, the order of each target candidate push information in the sequence is fixed, and only after the strategy with a higher priority selects the corresponding target candidate push information with the first proportion, the strategy with the second highest priority will select the corresponding target candidate push information with the first proportion, until the last target candidate push information in the sequence. According to the method of the present application, in the sequence, the target push strategy corresponding to the target candidate push information of each sequence number is randomly determined, except that the proportion of the target candidate push information selected by each target push strategy is still the corresponding first proportion. Since the strategy corresponding to the target candidate push information of each sequence number is determined completely randomly, the problem of fairness defects is avoided.

[0126] S104: Based on the target push strategy corresponding to each target push information to be selected and according to the push ranking result corresponding to the target push strategy, select target candidate push information from each initial candidate push information.

[0127] The embodiment of the present application first randomly determines the target push strategy corresponding to each target push information to be selected, and after determining the target push strategy, selects the target candidate push information from each initial candidate push information based on the ranking result of the target push strategy. For example, if the target push strategy of the first target candidate push information is determined to be strategy n, and the target push strategy of the second target candidate push information is determined to be strategy m, then the first initial candidate push information in the push ranking result obtained based on strategy n is used as the first target candidate push information, and the first initial candidate push information in the push ranking result obtained based on strategy mn is used as the second target candidate push information, until all the target candidate push information is obtained.

[0128] S105: Determine target push information from candidate target push information, and push the target push information to the target object.

[0129] It should be noted that the process of obtaining the target candidate push information through the above steps can immediately be the recall and rough sorting process in the information recommendation process, and the results of the rough sorting can be further refined to obtain the target push information, and the target push information can be pushed to the target object. In the refined sorting stage, the embodiment of the present application can use the ECPM strategy to balance the system benefits and object experience.

[0130] The embodiment of the present application obtains a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, wherein the candidate push information set includes at least one initial candidate push information, determines the push ranking result of each initial candidate push information in the candidate push information set under each push strategy, and randomly determines the target push strategy corresponding to each target candidate push information to be selected from at least one push strategy according to the first proportion. On the one hand, it ensures diversity when obtaining target candidate push information, and on the other hand, it solves the problem of lack of fairness in policy priority, ensures fair competition among policies, and is conducive to objective and efficient evaluation of policy effects. It saves time for troubleshooting policy coverage issues when policy priorities are repeatedly aligned and policy effects do not meet expectations, and also provides a unified effect evaluation mechanism at the framework level for long-term assessment of policy effects, thereby further improving the optimization efficiency of the push system.

[0131] Based on the above embodiments, as an optional embodiment, according to the first proportion, a target push strategy corresponding to each target candidate push information to be selected is randomly determined from at least one push strategy, including:

[0132] S201. Determine the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy.

[0133] It should be noted that, when the total amount of target candidate push information is determined, the amount of target candidate push information corresponding to each push strategy can be determined by multiplying the first proportion of each push strategy by the total amount (ie, the preset amount).

[0134] S202: Initialize a target candidate push information sequence, where each element in the target candidate push information sequence represents a target candidate push information to be selected.

[0135] The present embodiment does not specifically limit the order of steps S201 and S202, for example, S201 and S202 can be executed simultaneously. Each element in the initialized target candidate push information sequence is empty, but the number of elements is a preset number, that is, the same as the number of target candidate push information.

[0136] S203. According to the number of target candidate push information corresponding to each push strategy, randomly select a corresponding number of elements in the target candidate push information sequence, and mark the corresponding push strategies. The marked push strategies represent the target push strategies corresponding to the corresponding elements.

[0137] In the embodiment of the present application, when the push strategy is uploaded to the system, the system may generate a corresponding unique identifier, and when marking the element, the unique identifier of the target push strategy may be specifically marked.

[0138] It should be noted that the embodiment of the present application first determines the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy, and then initializes the target candidate push information sequence. Each element in the target candidate push information sequence represents a target candidate push information to be selected, that is, corresponds to a push strategy. Further, according to the number of target candidate push information corresponding to each push strategy, a corresponding number of elements are randomly selected in the target candidate push information sequence, and the corresponding push strategies are marked. This makes the target push strategies corresponding to each element in the target candidate push information sequence of the itinerary random, avoiding the fairness problem caused by the policy priority of the prior art, and also ensures that the number of strategies corresponding to each target candidate push information in the target candidate push information sequence can also meet the first proportion.

[0139] See also Figure 5, which exemplarily shows a flow chart of randomly determining the target push strategy corresponding to each target candidate push information to be selected in an embodiment of the present application. As shown in the figure, the embodiment of the present application can first determine the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy. In the figure, the number of target candidate push information corresponding to strategy one is 3, and the number of target candidate push information corresponding to strategy 2 is 2. The target candidate push information sequence is initialized. Each element in the target candidate push information sequence represents a target candidate push information to be selected. The initialized target candidate push information sequence in the figure has a total of 5 elements. First, (randomly or sequentially) 3 of the 5 elements are matched with strategy one, and then the remaining 2 elements are matched with strategy two. In the figure, the first 3 elements are marked as strategy one, and the last two elements are marked as strategy two. Then, the positions of the elements in the initialized target candidate push information sequence are randomly rearranged to obtain the randomly rearranged target candidate push information sequence. It can be seen from the figure that the positions of the elements corresponding to each strategy have been scattered.

[0140] Based on the above embodiments, as an optional embodiment, selecting target candidate push information from each initial candidate push information according to the push ranking result corresponding to the target push strategy includes:

[0141] S301: Determine reference information, where the reference information is selected target candidate push information.

[0142] It should be noted that the embodiment of the present application can determine the target candidate push information corresponding to each element one by one according to the order of the elements in the target candidate push information sequence, so that when determining the target candidate push information corresponding to each (current) element, the target push information selected by each element before the current element is used as reference information. For example, when determining the corresponding target candidate push information for the kth element in the target candidate push information sequence, all the target candidate push information corresponding to the 1st to k-1st elements are used as reference information.

[0143] S302: According to the push ranking result corresponding to the target push strategy, select the target candidate push information corresponding to the current operation from the information other than the reference information in each candidate push information.

[0144] After determining the target push strategy, an embodiment of the present application can determine the target candidate push information corresponding to the current operation from the push sorting results of the target push strategy and from the information other than the reference information. For example, the first information other than the reference information is determined as the target candidate push information corresponding to the current operation.

[0145] Specifically, the embodiment of the present application uses the target candidate push information that was last determined based on the target policy corresponding to the target candidate push information that is currently being selected as the benchmark information for the target candidate push information that is currently being selected, and starts from the first candidate push information that is located after the benchmark information in the push sorting result of the target policy, and compares them with the reference information in sequence, and uses the first candidate push information that is different from the reference information as the target candidate push information corresponding to the target element, and uses the currently selected target candidate push information as the target candidate push information determined this time based on the target policy.

[0146] It should be noted that when determining the target candidate push information, the present application uses the selected target candidate information as reference information and selects the target candidate push information from the information other than the reference information in the target sorting result to avoid selecting duplicate target candidate push information, thereby ensuring the diversity of the selected target candidate push information.

[0147] The following is a specific example to illustrate the process of the information push method of the embodiment of the present application. Figure 6 In this embodiment, the initial candidate push information set includes Ad1~Adt, and the push strategies include strategy one and strategy two, among which the first proportion of strategy one is 66.7%, and the first proportion of strategy two is 33.3%. The number of target candidate push information finally determined is 6.

[0148] According to the first proportion of the two push strategies, it is determined that the number of target candidate push information corresponding to strategy one is 4, and the number of target candidate push information corresponding to strategy two is 2. The target candidate push information sequence is initialized. The target candidate push information sequence has a total of 6 elements. According to the number of target candidate push information corresponding to the two push strategies, the first 4 elements are marked in order to correspond to strategy two, and the last 2 elements correspond to strategy two. Then, the elements of the target candidate push information sequence are randomly rearranged to obtain a randomly rearranged target candidate push information sequence, and the push ranking results of the initial candidate push information under each strategy are obtained respectively.

[0149] Since the first element in the target candidate push information sequence marks strategy one, the target candidate push information is determined from the push ranking result of strategy one, and the initial candidate push information Ad1 ranked 1 is used as the first target candidate push information in the target candidate push information sequence, and Ad1 is used as the target candidate push information determined this time based on strategy one.

[0150] Continue to obtain the second target candidate push information in the target candidate push information sequence. Since the push strategy corresponding to the target candidate push information is strategy 2, no target candidate push information has been selected in the push ranking result of strategy 2, and the initial candidate push information Ad4 ranked 1 is also different from Ad1, so Ad4 is used as the second target candidate push information in the target candidate push information sequence.

[0151] Continue to obtain the third target candidate push information in the target candidate push information sequence. Since the push strategy corresponding to the target candidate push information is strategy one, and the previous target candidate push information obtained based on strategy one is Ad1, then start looking for initial candidate push information different from the currently obtained target candidate push information - Ad1 and Ad4 - after Ad1. Since Ad3 ranked 2 is different from Ad1 and Ad4, Ad3 is used as the third target candidate push information in the target candidate push information sequence, and Ad3 is also used as the target candidate push information determined based on strategy one this time.

[0152] Similarly, the final target candidate push information queue can be obtained: Ad1, Ad4, Ad3, Ad7, Ad2, Ad6.

[0153] Based on the above embodiments, as an optional embodiment, obtaining at least one push strategy corresponding to the candidate push information set and a first proportion corresponding to each push strategy includes:

[0154] S401. Obtain diversified configuration information, where the diversified configuration information includes at least one policy group and a second proportion of each policy group, and each policy group includes at least one push policy.

[0155] It should be noted that the second proportion in the embodiment of the present application represents the proportion of the number of target candidate push information corresponding to the corresponding strategy group to the preset number. Therefore, it can be seen that the sum of the second proportions of all strategy groups should be 100%.

[0156] S402: Determine the third proportion corresponding to each push strategy in each strategy group.

[0157] The third proportion of the embodiment of the present application is used to characterize the proportion of the target candidate push information selected by the corresponding push strategy to the target candidate push information selected by all push strategies in the strategy group where the corresponding push strategy is located. It can be seen that the sum of the third proportions of all push strategies in a strategy group should be 100%.

[0158] See also Figure 7, which exemplarily shows a schematic diagram of the diversified configuration information of the embodiment of the present application. As shown in the figure, a diversified configuration information includes N policy groups, N is a natural number greater than 1, each policy group can set a second proportion, and the sum of all second proportions is equal to 100%. Each policy group includes n push policies, and the sum of the third proportions of the push policies under each policy group is equal to 100%. In addition, policies can be freely added or deleted under each policy group, and the number of added policies is at least 1.

[0159] S403: For each push policy, determine a first proportion of the push policy according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group.

[0160] For each push strategy, first, the push strategy appears in several strategy groups, then determine the second proportion of the strategy group where the push strategy is located, and multiply it by the third proportion of the push strategy in the corresponding strategy group to obtain the first proportion. The specific formula is:

[0161] U i1 =∑U j2 U ij3

[0162] Among them, U i1 represents the first proportion of strategy i, U j2 represents the second proportion of strategy group j where strategy i belongs, U ij3 Indicates the third proportion of strategy i in strategy group j.

[0163] For example, if there are two policy groups, the second proportion of policy group 1 is 80%, the second proportion of policy group 2 is 20%, the third proportion of policy 1 in policy group 1 is 20%, and the third proportion of policy 1 in policy group 2 is 50%, then the first proportion of policy 1 is: 80%×20%+20%×50%=26%.

[0164] It should be noted that since the first proportion of the embodiment of the present application is obtained through the second proportion of each policy group and the third proportion corresponding to each push policy in each policy group, the proportion information (including the second proportion and the third proportion) can be dynamically configured to adjust the number of policies corresponding to the final target candidate push information.

[0165] Based on the above embodiments, as an optional embodiment, determining the third proportion of policies in each policy group includes:

[0166] S501, obtaining the object information of the target object to which the information is to be pushed;

[0167] It should be emphasized that when the embodiments of the present application are applied to specific products or technologies, the object's permission or consent must be obtained, and the collection, use and processing of relevant data (including object information) must comply with relevant laws, regulations and standards of relevant countries and regions.

[0168] S502: Determine a third proportion corresponding to the object information according to the object information.

[0169] In the embodiment of the present application, the third proportion of the strategy is determined based on the object information of the target object, so that for different target objects, even if the same strategy group is used, push strategies with different first proportions can be obtained, laying the foundation for personalized push information.

[0170] Based on the above embodiments, as an optional embodiment, the object information includes feedback information for historical push information; for example, it may be the click rate, conversion rate, etc. of the target object for historical push information.

[0171] Accordingly, the embodiment of the present application determines the third proportion corresponding to the object information, including:

[0172] Determining a third proportion corresponding to the feedback information of the target object from the predetermined proportion configuration information;

[0173] The proportion configuration information includes a proportion relationship corresponding to at least one feedback information in a one-to-one manner; the proportion relationship is used to characterize the third proportion of each strategy in the strategy group.

[0174] Taking the feedback information as click rate as an example, the proportion configuration information of the embodiment of the present application can set the proportion relationship corresponding to multiple click rate intervals, so that after determining the actual click rate of the target object, the corresponding proportion relationship is determined according to the interval in which the actual click rate is located.

[0175] See also Figure 8 , which exemplarily shows a flow chart of an information push method according to another embodiment of the present application. As shown in the figure, the embodiment of the present application sets four gears for the feedback information of the object - click-through rate, and sets a proportion relationship for each gear respectively. It should be noted that, in the present embodiment, a strategy group includes three push strategies, so the proportion relationship is used to describe the third proportion relationship of the three push strategies.

[0176] By obtaining an information push request for a target object, the request includes information about the click-through rate of the target object or the gear corresponding to the click-through rate. It can be understood that if the request only includes the click-through rate, it is necessary to further determine the corresponding gear according to the click-through rate. First, determine whether the gear is less than 2. If the gear is not less than 2, that is, the gear is 3, then the proportion relationship corresponding to gear 3 is determined to be the third proportion corresponding to the object information, which is 10:0:0 in the figure, that is, when the strategy group is used to select the target candidate push information, the first strategy in the strategy group is used to select the target candidate push information. If it is determined that the gear is less than 2, it is further determined whether the gear is less than 1. If it is less than 1, it means that the gear is 0, and the third proportion of the strategy is determined according to the proportion relationship corresponding to gear 0; if it is not less than 1, it means that the gear is 1 or 2. For the purpose of convenience of description, the gears 1 and 2 in this application correspond to the same proportion relationship, which shows that a proportion relationship in an embodiment of the present application can correspond to multiple gears (that is, multiple feedback information).

[0177] After determining the third proportion of the strategy, the first proportion of the strategy can be obtained by combining the second proportion of the strategy group, and then the push sorting results of each strategy are randomly cross-sorted according to the first proportion through the above embodiment to obtain a target candidate push information queue, and the target candidate push information is finely sorted to finally obtain the target push information, and the target push information is pushed to the target object for exposure.

[0178] Based on the above embodiments, as an optional embodiment, obtaining the diversified configuration information also includes the step of dynamically adjusting the third proportion of the strategy, which specifically includes:

[0179] S601: Create at least one policy group, and initialize the third proportion of each push policy in the policy group.

[0180] In an embodiment of the present application, after registering a policy, the policy party can create at least one policy group according to preset rules based on the policies already registered on the system. For example, push policies for the same type of push information can be created as a policy group. Specifically, if policies 1 and 2 are push policies for academic type push information, and policies 3 and 4 are push policies for game type push information, then policies 1 and 2 can be set in one policy group, and policies 3 and 4 can be set in another policy group.

[0181] In the embodiment of the present application, when initializing the third proportion of the push strategy, it can be set based on the historical effect of the push strategy. If the push strategy has no historical effect yet, it can be set randomly within a preset range. It can be understood that the sum of the initialized third proportions of all push strategies in a strategy group is also 100%.

[0182] S602: In response to an update of a third proportion of at least one push policy in a policy group or an update of the number of push policies in a policy group, determine a third proportion of all push policies in an updated policy group.

[0183] The embodiment of the present application supports updating the number of push strategies in the strategy group, and can also update the third proportion of the push strategies, which can further improve the adaptability when pushing information and improve the optimization efficiency of the push system. Through the dynamic quota mechanism, the push system is allowed to customize the sorting strategy combination with different optimization goals for different objects, provide more personalized recommendation results, improve the object experience, and thus build a long-term healthy push ecosystem.

[0184] S603. If the sum of the third proportions of all push policies in the updated policy group is less than the preset value, increase the third proportion of at least one push policy in the policy group until the third proportion of all push policies in the policy group is equal to the preset value; if the sum of the third proportions of all push policies in the updated policy group is greater than the preset value, compress the updated third proportion of at least one push policy in the policy group until the third proportions of all push policies in the policy group are equal to the preset value.

[0185] It should be understood that the preset value in the present application can be 100%. If the number of push policies or the third ratio of push policies is updated, resulting in the sum of the third ratios of all push policies in the updated policy group not being equal to the preset value, it is necessary to adjust the third ratio of at least one push policy so that the third ratios of all push policies in the policy group are equal to the preset value.

[0186] See also Fig. 9 , which exemplarily shows a flow chart of an information push method according to another embodiment of the present application, as shown in the figure, including:

[0187] S701, the policy party registers and pushes the policy;

[0188] S702: Create diversified configuration information according to the push policy registered by at least one policy party, where the diversified configuration information includes at least one policy group and a second proportion of each policy group, and each policy group includes at least one push policy;

[0189] S703: Determine the third proportion of each push strategy in each strategy group and a candidate push information set according to the object information of the target object;

[0190] S704: For each push policy, determine a first proportion of the push policy according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group;

[0191] S705, determining the push ranking result of each initial candidate push information in the candidate push information set under each push strategy;

[0192] S706: Determine the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy;

[0193] S707, initializing a target candidate push information sequence, wherein each element in the target candidate push information sequence represents a target candidate push information to be selected;

[0194] S708. According to the number of target candidate push information corresponding to each push strategy, randomly select a corresponding number of elements in the target candidate push information sequence, and mark the corresponding push strategies, wherein the marked push strategies represent the target push strategies corresponding to the corresponding elements;

[0195] S709, determining reference information, where the reference information is the selected target candidate push information;

[0196] S710, selecting target candidate push information corresponding to the current operation from information other than reference information in each candidate push information according to the push ranking result corresponding to the target push strategy;

[0197] S711. Determine target push information from candidate target push information, and push the target push information to the target object.

[0198] The present application embodiment provides an information push device, such as Fig.10 As shown, the information push device may include: an information acquisition module 101, a single strategy sorting module 102, a strategy mixing module 103, a push information selection module 104 and a push module 105, wherein:

[0199] The information acquisition module 101 is used to acquire a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, wherein the candidate push information set includes at least one initial candidate push information; the first proportion is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to a preset number;

[0200] The single-strategy sorting module 102 determines the push sorting results of each initial candidate push information in the candidate push information set under each push strategy;

[0201] A strategy shuffling module 103 is used to randomly determine a target push strategy corresponding to each target candidate push information to be selected from at least one push strategy according to the first proportion;

[0202] The push information selection module 104 is used to select target candidate push information from each initial candidate push information based on the target push strategy corresponding to each target push information to be selected and according to the push sorting result corresponding to the target push strategy;

[0203] The push module 105 is used to determine target push information from candidate target push information and push the target push information to the target object.

[0204] The device of the embodiments of the present application can execute the method provided by the embodiments of the present application, and the implementation principles are similar. The actions performed by each module in the device of each embodiment of the present application correspond to the steps in the method of each embodiment of the present application. For the detailed functional description of each module of the device, please refer to the description in the corresponding method shown in the previous text, which will not be repeated here.

[0205] In an embodiment of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory. The processor executes the above-mentioned computer program to implement the steps of the information push method. Compared with the related art, the following can be achieved:

[0206] By obtaining a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, the candidate push information set includes at least one initial candidate push information, and the push ranking results of each initial candidate push information in the candidate push information set under each push strategy are determined. According to the first proportion, a target push strategy corresponding to each target candidate push information to be selected is randomly determined from at least one push strategy. On the one hand, diversity is ensured when obtaining target candidate push information, and on the other hand, the problem of lack of fairness in strategy priority is solved, and fair competition among strategies is ensured, which is conducive to objective and efficient evaluation of strategy effects. It not only saves time for troubleshooting strategy coverage problems when strategy priorities are repeatedly aligned and strategy effects do not meet expectations, but also provides a unified effect evaluation mechanism at the framework level for long-term assessment of strategy effects, thereby further improving the optimization efficiency of the push system.

[0207] As an optional embodiment, the policy shuffling module includes:

[0208] An information quantity determination submodule, used to determine the quantity of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy;

[0209] A sequence initialization submodule is used to initialize a target candidate push information sequence, wherein each element in the target candidate push information sequence represents a target candidate push information to be selected;

[0210] The random marking submodule is used to randomly select a corresponding number of elements in the target candidate push information sequence according to the number of target candidate push information corresponding to each push strategy, and mark the corresponding push strategy. The marked push strategy represents the target push strategy corresponding to the corresponding element.

[0211] As an optional embodiment, the push information selection module includes:

[0212] The reference information determination submodule is used to determine the reference information, where the reference information is the selected target candidate push information;

[0213] The sorting and filtering submodule is used to select the target candidate push information corresponding to the current operation from the information other than the reference information in each candidate push information according to the push sorting result corresponding to the target push strategy.

[0214] As an optional embodiment, the information acquisition module includes:

[0215] A configuration information acquisition submodule is used to acquire diversified configuration information, wherein the diversified configuration information includes at least one policy group and a second proportion of each policy group, each policy group includes at least one push policy, and the second proportion represents the proportion of the number of target candidate push information corresponding to the corresponding policy group to the preset number;

[0216] A third proportion determination submodule is used to determine a third proportion corresponding to each push strategy in each strategy group, where the third proportion is used to represent the proportion of the target candidate push information selected by the corresponding push strategy to the target candidate push information selected by all push strategies in the strategy group where the corresponding push strategy belongs;

[0217] The first proportion determination submodule is used to determine, for each push policy, a first proportion of the push policy according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group.

[0218] As an optional embodiment, the object information includes feedback information for historical push information;

[0219] The third proportion determination submodule is specifically used to: determine a third proportion corresponding to the feedback information of the target object from the predetermined proportion configuration information;

[0220] The proportion configuration information includes a proportion relationship corresponding to at least one feedback information in a one-to-one manner; the proportion relationship is used to characterize the third proportion of each strategy in the strategy group.

[0221] As an optional embodiment, the push device further includes a configuration update module, and the configuration update module includes:

[0222] A policy group creation submodule is used to create at least one policy group and initialize the third proportion of each push policy in the policy group;

[0223] A policy update submodule, configured to determine the third proportions of all push policies in the updated policy group in response to an update of the third proportion of at least one push policy in the policy group or an update of the number of push policies in the policy group;

[0224] A first proportion adjustment submodule is used to increase the third proportion of at least one push policy in the policy group if the sum of the third proportions of all push policies in the updated policy group is less than a preset value, until the third proportions of all push policies in the policy group are equal to the preset value;

[0225] The second proportion adjustment submodule compresses the updated third proportion of at least one push policy in the policy group until the third proportions of all push policies in the policy group are equal to the preset value if the sum of the third proportions of all push policies in the policy group after the update is greater than the preset value.

[0226] In an alternative embodiment, an electronic device is provided, such as Fig.11 As shown, Fig.11 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0227] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0228] The bus 4002 may include a path to transmit information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0229] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0230] The memory 4003 is used to store a computer program for executing the embodiment of the present application, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the aforementioned method embodiment, by obtaining a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, the candidate push information set includes at least one initial candidate push information, and the push sorting result of each initial candidate push information in the candidate push information set under each push strategy is determined. According to the first proportion, the target push strategy corresponding to each target candidate push information to be selected is randomly determined from at least one push strategy, on the one hand, the diversity when obtaining the target candidate push information is guaranteed, and on the other hand, the problem of lack of fairness in the strategy priority is solved, and fair competition between the strategies is guaranteed, which is conducive to objectively and efficiently evaluating the strategy effect, saving the time of troubleshooting the strategy coverage problem when the strategy priority is repeatedly aligned and the strategy effect does not meet the expectation, and also providing a unified effect evaluation mechanism at the framework level for long-term assessment of the effect of the strategy, and further improving the optimization efficiency of the push system.

[0231] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented, by obtaining a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, the candidate push information set includes at least one initial candidate push information, and the push ranking result of each initial candidate push information in the candidate push information set under each push strategy is determined. According to the first proportion, a target push strategy corresponding to each target candidate push information to be selected is randomly determined from at least one push strategy. On the one hand, diversity is ensured when obtaining target candidate push information, and on the other hand, the problem of lack of fairness in policy priority is solved, and fair competition among policies is ensured, which is conducive to objective and efficient evaluation of policy effects. It not only saves time for troubleshooting policy coverage problems when policy priorities are repeatedly aligned and policy effects do not meet expectations, but also provides a unified effect evaluation mechanism at the framework level for long-term assessment of policy effects, thereby further improving the optimization efficiency of the push system.

[0232] The embodiment of the present application also provides a computer program product, including a computer program, which can implement the steps and corresponding contents of the aforementioned method embodiment when executed by a processor.

[0233] The terms "first", "second", "third", "fourth", "1", "2", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than that shown or described in the drawings.

[0234] It should be understood that, although each operation step is indicated by arrows in the flowchart of the embodiment of the present application, the implementation order of these steps is not limited to the order indicated by the arrows. Unless clearly stated herein, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders according to demand. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on actual implementation scenarios. Some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage in these sub-steps or stages may also be executed at different times respectively. In different scenarios of execution time, the execution order of these sub-steps or stages may be flexibly configured according to demand, and the embodiment of the present application does not limit this.

[0235] The above are only optional implementation methods for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the scheme of the present application, other similar implementation methods based on the technical ideas of the present application are also within the protection scope of the embodiments of the present application.

Claims

1. An information push method, characterized in that: include: Acquire a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, wherein the candidate push information set includes at least one initial candidate push information; The first proportion is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to the preset number; Determine a push ranking result of each initial candidate push information in the candidate push information set under each push strategy; According to the first proportion, randomly determine the target push strategy corresponding to each target candidate push information to be selected from the at least one push strategy, so that the ratio of the target candidate push information determined based on each push strategy to the preset number is the first proportion corresponding to the corresponding push strategy, and the order of determining the target candidate push information based on different push strategies is random; Based on the target push strategy corresponding to each target push information to be selected, and according to the push ranking result corresponding to the target push strategy, select the target candidate push information from the initial candidate push information; Determine target push information from the target candidate push information, and push the target push information to the target object; The step of selecting target candidate push information from the initial candidate push information according to the push ranking result corresponding to the target push strategy includes: Determine reference information, where the reference information is the selected target candidate push information; Take the target candidate push information determined last time based on the target policy corresponding to the currently selected target candidate push information as the benchmark information, and start from the first candidate push information after the benchmark information in the push sorting result of the target policy, and compare them with the reference information in sequence, and take the first candidate push information different from the reference information as the target candidate push information corresponding to the target element, and take the currently selected target candidate push information as the target candidate push information determined this time based on the target policy.

2. The information push method according to claim 1, characterized in that: The step of randomly determining, according to the first proportion, a target push strategy corresponding to each target candidate push information to be selected from the at least one push strategy includes: Determine the number of target candidate push information corresponding to each push strategy according to the first proportion of each push strategy; Initializing a target candidate push information sequence, wherein each element in the target candidate push information sequence represents a target candidate push information to be selected; According to the number of target candidate push information corresponding to each push strategy, a corresponding number of elements are randomly selected from the target candidate push information sequence, and the corresponding push strategies are marked. The marked push strategies represent the target push strategies corresponding to the corresponding elements.

3. The method according to claim 1, characterized in that The obtaining of at least one push strategy corresponding to the candidate push information set and a first proportion corresponding to each push strategy includes: Obtaining diversified configuration information, the diversified configuration information including at least one policy group and a second proportion of each policy group, each policy group including at least one push policy, the second proportion representing the proportion of the number of target candidate push information corresponding to the corresponding policy group to the preset number; Determine a third proportion corresponding to each push policy in each policy group, wherein the third proportion is used to represent the proportion of the target candidate push information selected by the corresponding push policy to the target candidate push information selected by all push policies in the policy group where the corresponding push policy belongs; For each of the push policies, the first proportion of the push policy is determined according to the second proportion of the policy group to which the push policy belongs and the third proportion of the push policy in the corresponding policy group.

4. The method according to claim 3, characterized in that The determining the third proportion of the strategies in each strategy group includes: Get the object information of the target object of the information to be pushed; A third proportion corresponding to the object information is determined according to the object information.

5. The method according to claim 4, characterized in that The object information includes feedback information for historical push information; The determining a third proportion corresponding to the object information includes: Determining a third proportion corresponding to the feedback information of the target object from predetermined proportion configuration information; The proportion configuration information includes a proportion relationship corresponding to at least one type of feedback information in a one-to-one manner; the proportion relationship is used to characterize the third proportion of each strategy in the strategy group.

6. The method according to claim 3, characterized in that The obtaining of the diversified configuration information also includes: Create at least one policy group, and initialize the third proportion of each push policy in the policy group; In response to an update of a third proportion of at least one push policy in the policy group or an update of the number of push policies in the policy group, determining an updated third proportion of all push policies in the policy group; If the sum of the third proportions of all push policies in the updated policy group is less than a preset value, increasing the third proportion of at least one push policy in the policy group until the third proportions of all push policies in the policy group are equal to the preset value; If the sum of the third proportions of all push policies in the policy group after the update is greater than the preset value, the updated third proportion of at least one push policy in the policy group is compressed until the third proportions of all push policies in the policy group are equal to the preset value.

7. An information push device, characterized in that: include: An information acquisition module, used to acquire a candidate push information set, at least one push strategy corresponding to the candidate push information set, and a first proportion corresponding to each push strategy, wherein the candidate push information set includes at least one initial candidate push information; The first proportion is the proportion of the number of target candidate push information to be selected corresponding to the corresponding push strategy to the preset number; A single strategy sorting module determines a push sorting result of each initial candidate push information in the candidate push information set under each push strategy; a strategy shuffling module, configured to randomly determine, from the at least one push strategy, a target push strategy corresponding to each target candidate push information to be selected, according to the first proportion, so that the target candidate push information determined based on each push strategy accounts for a preset number of the first proportion corresponding to the corresponding push strategy, and the order of determining the target candidate push information based on different push strategies is random; A push information selection module, configured to select target candidate push information from the initial candidate push information based on the target push strategy corresponding to each target push information to be selected and according to the push sorting result corresponding to the target push strategy; A push module, used to determine target push information from the target candidate push information, and push the target push information to the target object; The push information selection module is specifically used for: Determine reference information, where the reference information is the selected target candidate push information; Take the target candidate push information determined last time based on the target policy corresponding to the currently selected target candidate push information as the benchmark information, and start from the first candidate push information after the benchmark information in the push sorting result of the target policy, and compare them with the reference information in sequence, and take the first candidate push information different from the reference information as the target candidate push information corresponding to the target element, and take the currently selected target candidate push information as the target candidate push information determined this time based on the target policy.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the information pushing method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the information push method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the information pushing method according to any one of claims 1 to 6 are implemented.

Images