Exposure recommendation method and system, electronic equipment and storage medium
By using the conversion rate model and exposure value model to estimate the probability of product conversion and combining business strategies, the problem of insufficient single exposure method in the existing technology is solved, more accurate and effective recommendation services are achieved, and user experience is improved.
Patent Information
- Application Number
- CN202411467770.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-02-18
AI Technical Summary
The single product exposure method in the prior art is not enough to provide recommendation services accurately and effectively, and cannot meet users' product exposure needs at different stages, resulting in a decline in user experience.
By collecting and processing basic data of users and products, using conversion rate models and exposure value models to estimate the conversion probability of products, perform weighted fusion to calculate product scores, and determine recommended products in combination with exposure position and business control strategies.
It achieves more effective and accurate recommendation services, meets users' product exposure needs at different stages, and improves user experience.
Smart Images
Figure CN119477455A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to an exposure recommendation method, system, electronic device and storage medium. Background Art
[0002] With the development of the Internet, recommendation systems, as the main scenario of the Internet, have been affecting the lives of Internet users all the time. In various application scenarios of recommendation systems, recommendation models are the core components of recommendation systems. Recommendation systems recommend different products to different users through recommendation models, which directly affects the user experience of recommendation services. In recent years, recommendation model technology has developed rapidly, and research on recommendation model technology has become an important branch of artificial intelligence (AI) research.
[0003] In the industry, the main way for recommendation models to interact with users is through product exposure. By analyzing user behavior, we found that users often need to go through several stages, such as first contact, interest, deepening impression, desire to consume, and consumption. This cannot be satisfied by a one-time product exposure method. Moreover, when a user is clearly not interested in a product, recommending it will often only backfire, making the user feel bored and even doubt the recommendation ability of the entire platform. Therefore, an accurate and effective product exposure method is the most important step in interacting with users.
[0004] At present, traditional exposure methods can be roughly divided into: 1) Feed (the smallest unit of information aggregation) flow-based exposure methods, which tend to be a one-time exposure method for products; 2) E-commerce-based exposure methods, which tend to be a repetitive exposure method. However, at this stage, a single product exposure method is no longer sufficient to provide accurate and effective recommendation services. Summary of the invention
[0005] The present invention provides an exposure recommendation method, system, electronic device and storage medium, which are used to solve the defect that a single product exposure method in the prior art is not sufficient to accurately and effectively provide recommendation services, and achieve the purpose of providing recommendation services more effectively and accurately.
[0006] The present invention provides an exposure recommendation method, comprising: In response to an access request sent by a target user, original basic data is collected and processed to obtain target basic data; the original basic data includes original user data and original product data; Perform product recall based on the target basic data to obtain a candidate product set; For each candidate product in the candidate product set, a first conversion probability of the candidate product for the target user is estimated using a conversion rate model, and a second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is estimated using an exposure value model; Performing a weighted fusion of the first conversion probability and the second conversion probability to obtain a score of the candidate product; Based on the scores of the candidate products, combined with the exposure position and the business control strategy, the recommended products to be exposed to the target users are determined and displayed.
[0007] According to an exposure recommendation method provided by the present invention, the exposure value model is trained by the following steps: Input the historical exposure information of different products on the product side for different users, the conversion rates after different historical exposure times, and the historical exposure sequences of different users on the user side who exposed different products at different times into the recurrent neural network, and predict the conversion probability after different users are exposed to different products this time; the historical exposure information includes at least one of the historical exposure times, historical exposure duration, historical exposure interval, and other products that were exposed at the same time in history; The recurrent neural network is trained based on each of the predicted conversion probabilities and the corresponding actual conversion results until the model converges to obtain the exposure value model.
[0008] According to an exposure recommendation method provided by the present invention, the method of using an exposure value model to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time includes: The historical exposure information of the candidate product on the product side for the target user, the conversion rate after different historical exposure times, and the historical exposure sequence of the target user on the user side when exposing the candidate product at different times are input into the exposure value model, and the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is output.
[0009] According to an exposure recommendation method provided by the present invention, based on the scores of the candidate products, combined with the exposure position and the business control strategy, the product recommended for exposure to the target user is determined and displayed, including: Determining exposure restrictions for different products based on the business control strategy; Determine the displayable quantity of the product based on the exposure position; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products; The product recommended for exposure to the target user this time is displayed at the exposure position.
[0010] According to an exposure recommendation method provided by the present invention, the restricted exposure conditions of different products include that the exposure time interval of the first product is a preset time interval; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the exposure time interval of the first product corresponding to the score of the target quantity ranked higher in the sorting result is not the preset time interval, the score of the first product is removed from the sorting result, and the product corresponding to the score of the target quantity ranked higher is reselected as the product recommended for exposure to the target user this time; the target quantity is equal to the displayable quantity of the product.
[0011] According to an exposure recommendation method provided by the present invention, the limited exposure conditions of different products include that the maximum exposure duration of the second product is a preset duration; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the exposure time of the second product corresponding to the score of the target number ranked at the top in the sorting result reaches the preset time, the score of the second product will be removed from the sorting result, and the product corresponding to the score of the target number ranked at the top will be reselected as the product recommended for exposure to the target user this time; the target number is equal to the displayable number of the product.
[0012] According to an exposure recommendation method provided by the present invention, the restricted exposure conditions of different products include limiting continuous exposure of a third product within a preset period; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the third product corresponding to the target number score ranked high in the sorting result is continuously exposed within the preset period, the score of the third product will be removed from the sorting result, and the product corresponding to the target number score ranked high will be reselected as the product recommended for exposure to the target user this time; the target number is equal to the displayable number of the product.
[0013] The present invention also provides an exposure recommendation system, comprising: The data layer is used to respond to the access request sent by the target user, collect the original basic data and process the original basic data to obtain the target basic data; the original basic data includes original user data and original product data; The model layer is used for: Perform product recall based on the target basic data to obtain a candidate product set; For each candidate product in the candidate product set, a first conversion probability of the candidate product for the target user is estimated using a conversion rate model, and a second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is estimated using an exposure value model; Performing a weighted fusion of the first conversion probability and the second conversion probability to obtain a score of the candidate product; The display layer is used to determine and display the recommended products to be exposed to the target user based on the scores of the candidate products, combined with the exposure position and business control strategy.
[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, an exposure recommendation method as described above is implemented.
[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the exposure recommendation method as described above is implemented.
[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the exposure recommendation method as described above is implemented.
[0017] The exposure recommendation method, system, electronic device and storage medium provided by the present invention first respond to the access request sent by the target user, collect the original basic data and process the original basic data to obtain the target basic data; the original basic data includes the original user data and the original product data; then, based on the target basic data, the product is recalled to obtain a candidate product set; then, for each candidate product in the candidate product set, the first conversion probability of the candidate product for the target user is estimated using a conversion rate model, which is biased towards the user's preference for different candidate products; the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is estimated using an exposure value model, which is biased towards the value of this exposure; The first conversion probability and the second conversion probability are weightedly fused to obtain the score of the candidate product; in this way, for the candidate products that the target user is still interested in, the first conversion probability and the second conversion probability are higher, and the score is also high; for the candidate products that the target user has lost interest in, the first conversion probability and the second conversion probability are lower, and the score is also low; finally, based on the score of each candidate product, combined with the exposure position and business control strategy, the product recommended to the target user this time is determined and displayed; in this way, the candidate products that the target user is still interested in can be continuously exposed; for the candidate products that the target user has lost interest in, their exposure opportunities are reduced to avoid wasting exposure positions, thereby providing recommendation services more effectively and accurately. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 It is a flowchart of the exposure recommendation method provided by an embodiment of the present invention.
[0020] Figure 2 Schematic diagram of the structure of the exposure recommendation system provided by an embodiment of the present invention.
[0021] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0023] Combine the following Figure 1 The exposure recommendation method of the present invention is described.
[0024] Please refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of the exposure recommendation method provided by an embodiment of the present invention. Figure 1 As shown, the method may include steps 101 to 105.
[0025] Step 101, responding to an access request sent by a target user, collecting original basic data and processing the original basic data to obtain target basic data; the original basic data includes original user data and original product data.
[0026] Specifically, the data layer is the guarantee of the data resources of the exposure recommendation system, and is mainly responsible for the collection and processing of basic data. The exposure recommendation system is a general analysis platform for various businesses. Therefore, the construction of the data layer plays a very important role in the design of the exposure recommendation system.
[0027] When the target user clicks on a page or link, an access request is sent to the data layer. The data layer responds to the access request and can collect and process the original basic data to obtain the target basic data.
[0028] Original basic data includes original user data and original product data. Among them, original user data includes basic user information (such as gender, age, geographic location, device used, network, etc.) and user historical behavior data (such as browsing history, click history, order history, purchase history and other feedback information). Based on basic user information and user historical behavior data, build user portraits. Original product data mainly includes basic product information (such as classifications at all levels, brands, attributes, descriptive words, product pictures, etc.) and user historical behavior data on products (such as positive feedback information such as browsing number, click number, order number, purchase number and negative feedback information such as refund, complaint, etc.). Based on basic product information and user historical behavior data on products, build product portraits.
[0029] The model layer is the core of the exposure recommendation system. The model layer is introduced below through steps 102 to 104.
[0030] Step 102: Perform product recall based on the target basic data to obtain a candidate product set.
[0031] Specifically, recall methods such as hot spot sorting, inverted sorting of labels at various levels, and collaborative filtering can be used to recall products based on user portraits and product portraits to obtain a set of candidate products.
[0032] Step 103: for each candidate product in the candidate product set, use the conversion rate model to estimate the first conversion probability of the candidate product for the target user, and use the exposure value model to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time.
[0033] Specifically, the conversion rate model can select a commonly used fine ranking model in the industry, such as LightGBM, DeepFM, etc., but the present embodiment is not limited thereto. Input the target basic data (such as user basic information, product basic information, and processed user historical behavior data) into the fine ranking model for training to obtain a conversion rate model. For each candidate product in the candidate product set, input the user basic information of the target user, the product basic information of the candidate product, and the processed historical behavior data of the target user into the conversion rate model, and the first conversion probability proba_A of the candidate product for the target user can be estimated.
[0034] In one embodiment, the exposure value model is trained by the following steps: Input the historical exposure information of different products on the product side for different users, the conversion rates after different historical exposure times (e.g., historical exposure 1 time, historical exposure 2 times, historical exposure 3 times and above), and the historical exposure sequence of different users on the user side exposing different products at different times (e.g., user 1 exposing product U_1 at time t_1, exposing product U_2 at time t_2, ..., exposing product U_k at time t_k) into a recurrent neural network (RNN) to predict the conversion probability after different users are exposed to different products this time; the historical exposure information includes at least one of the historical exposure times, historical exposure duration, historical exposure interval, and other products that were exposed at the same time in history; The recurrent neural network is trained based on each predicted conversion probability and the corresponding actual conversion results until the model converges to obtain an exposure value model.
[0035] In this embodiment, the recurrent neural network is trained by using the historical exposure information on the product side, the conversion rates after different historical exposure times, and the historical exposure sequences on the user side, so as to obtain an exposure value model that is biased towards the value of this exposure.
[0036] In one embodiment, the use of the exposure value model in step 103 to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time includes: inputting historical exposure information of the candidate product on the product side for the target user, the conversion rate after different historical exposure times (for example, historical exposure 1 time, historical exposure 2 times, historical exposure 3 times and above), and the historical exposure sequence of the target user on the user side when the candidate product is exposed at different times into the exposure value model, and outputting the second conversion probability proba_B of the candidate product for the target user after the candidate product is exposed to the target user this time.
[0037] In this embodiment, the historical exposure information of candidate products on the product side for target users, the conversion rates after different historical exposure times, and the historical exposure sequences of target users on the user side when they are exposed to candidate products at different times are input into the exposure value model, so that a second conversion probability that can reflect the value of this exposure can be estimated.
[0038] Step 104: Perform weighted fusion on the first conversion probability and the second conversion probability to obtain a score for the candidate product.
[0039] Specifically, the score of the candidate product can be calculated by score = proba_A^α * proba_B^ß. In which, score represents the score of the candidate product, proba_A represents the first conversion probability, proba_B represents the second conversion probability, α represents the importance weight of the conversion rate model in the recommendation, and ß represents the importance weight of the exposure value model in the recommendation.
[0040] Step 105: Based on the scores of the candidate products, combined with the exposure position and the business control strategy, determine the products recommended for exposure to the target user this time and display them.
[0041] Specifically, the display layer is responsible for determining the most suitable product for this exposure recommendation to the target user and displaying it. After calculating the scores of each candidate product, the display layer also controls the exposure time interval, exposure duration, and whether the product is continuously exposed, etc., in combination with the exposure position and business control strategy, in order to achieve a better user experience, to find the most suitable product for this exposure recommendation to the target user, and display it to the target user.
[0042] In one embodiment, based on the scores of each candidate product, combined with the exposure position and the business control strategy, the products recommended for exposure to the target user are determined and displayed, including: determining the exposure restrictions of different products based on the business control strategy; determining the displayable number of products based on the exposure position; under the constraints of the exposure restrictions of different products and the displayable number of products, filtering out the products recommended for exposure to the target user from the sorting results of the scores of each candidate product; displaying the products recommended for exposure to the target user at the exposure position can improve the user experience in combination with the business control strategy.
[0043] The following is a detailed description of the screening methods under different exposure restriction situations.
[0044] In a first possible implementation, the restricted exposure conditions of different products include that the exposure time interval of the first product is a preset time interval; under the constraints of the restricted exposure conditions of different products and the displayable quantity of the products, the products recommended for exposure to the target user this time are screened out from the sorting results of the scores of each candidate product, including: obtaining the sorting results of the scores of each candidate product from large to small; if the exposure time interval of the first product corresponding to the score of the target quantity ranked first in the sorting results is not the preset time interval, then the score of the first product is removed from the sorting results, and the product corresponding to the score of the target quantity ranked first is reselected as the product recommended for exposure to the target user this time; the target quantity is equal to the displayable quantity of the product.
[0045] In this implementation, it is possible to avoid displaying the first product whose exposure time interval is a preset time interval, thereby avoiding frequent exposure of the user to the same product, thereby improving the user experience.
[0046] In a second possible implementation, the restricted exposure conditions of different products include that the maximum exposure time of the second product is a preset time; under the constraints of the restricted exposure conditions of different products and the displayable quantity of the products, the products recommended for exposure to the target user this time are screened out from the sorting results of the scores of each candidate product, including: obtaining the sorting results of the scores of each candidate product from large to small; if the exposure time of the second product corresponding to the score of the target quantity ranked at the top in the sorting results reaches the preset time, the score of the second product is removed from the sorting results, and the product corresponding to the score of the target quantity ranked at the top is reselected as the product recommended for exposure to the target user this time; the target quantity is equal to the displayable quantity of the product.
[0047] In this implementation, it is possible to avoid the exposure time of the same product being too long, thereby preventing users from feeling bored, thereby improving the user experience.
[0048] In a third possible implementation, the restricted exposure conditions of different products include restricting continuous exposure of a third product within a preset period; under the constraints of the restricted exposure conditions of different products and the displayable quantity of products, the products recommended for exposure to the target user are screened out from the sorting results of the scores of each candidate product, including: obtaining the sorting results of the scores of each candidate product from large to small; if the third product corresponding to a target quantity score ranked high in the sorting results is continuously exposed within a preset period (for example, 1 minute), the score of the third product is removed from the sorting results, and the product corresponding to the target quantity score ranked high is reselected as the product recommended for exposure to the target user; the target quantity is equal to the displayable quantity of the product.
[0049] In this implementation, continuous exposure of the same product within a preset period can be avoided, thereby preventing users from feeling bored, thereby improving user experience.
[0050] The exposure recommendation method provided by the embodiment of the present invention first responds to the access request sent by the target user, collects the original basic data and processes the original basic data to obtain the target basic data; the original basic data includes the original user data and the original product data; then, based on the target basic data, the product is recalled to obtain a candidate product set; then, for each candidate product in the candidate product set, a conversion rate model is used to estimate the first conversion probability of the candidate product for the target user, which is biased towards the user's preference for different candidate products; an exposure value model is used to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time, which is biased towards the value of this exposure; the first conversion probability is converted into the value of the first conversion probability. The first conversion rate and the second conversion probability are weightedly integrated to obtain the score of the candidate product; in this way, for the candidate products that the target user is still interested in, the first conversion probability and the second conversion probability are higher, and the score is also high; for the candidate products that the target user has lost interest in, the first conversion probability and the second conversion probability are lower, and the score is also low; finally, based on the score of each candidate product, combined with the exposure position and business control strategy, the product recommended to the target user this time is determined and displayed; in this way, the candidate products that the target user is still interested in can be continuously exposed; for the candidate products that the target user has lost interest in, their exposure opportunities are reduced to avoid wasting exposure positions, so that recommendation services can be provided more effectively and accurately.
[0051] The exposure recommendation device provided by the present invention is described below. The exposure recommendation device described below and the exposure recommendation method described above can be referred to each other.
[0052] Please refer to Figure 2 , Figure 2 Schematic diagram of the structure of the exposure recommendation system provided by the embodiment of the present invention. Figure 2 As shown, the exposure recommendation system includes: Data layer 1 is used to respond to the access request sent by the target user, collect the original basic data and process the original basic data to obtain the target basic data; the original basic data includes original user data and original product data; Model layer 2 is used for: Conduct product recall based on target basic data to obtain a candidate product set; For each candidate product in the candidate product set, the first conversion probability of the candidate product for the target user is estimated using the conversion rate model, and the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user is estimated using the exposure value model; The first conversion probability and the second conversion probability are weighted and integrated to obtain the score of the candidate product; Display layer 3 is used to determine and display the recommended products to the target users based on the scores of each candidate product, combined with the exposure position and business control strategy.
[0053] In one embodiment, the model layer 2 is also used to: Input the historical exposure information of different products for different users on the product side, the conversion rates after different historical exposure times, and the historical exposure sequences of different users on the user side who were exposed to different products at different times into the recurrent neural network to predict the conversion probability after different users were exposed to different products this time; the historical exposure information includes at least one of the historical exposure times, historical exposure duration, historical exposure interval, and other products that were exposed at the same time in history; The recurrent neural network is trained based on each predicted conversion probability and the corresponding actual conversion results until the model converges to obtain an exposure value model.
[0054] In one embodiment, the model layer 2 is specifically used for: The historical exposure information of the candidate products on the product side for the target users, the conversion rates after different historical exposure times, and the historical exposure sequences of the target users on the user side who are exposed to the candidate products at different times are input into the exposure value model, and the second conversion probability of the candidate products for the target users after the candidate products are exposed to the target users this time is output.
[0055] In one embodiment, the display layer 3 is specifically used for: Determine the exposure restrictions for different products based on business control strategies; Determine the number of products that can be displayed based on the exposure position; Under the constraints of the limited exposure of different products and the number of products that can be displayed, select the products recommended for exposure to the target users from the ranking results of the scores of each candidate product; The products recommended to the target users will be displayed in the exposure position.
[0056] In one embodiment, the restricted exposure conditions of different products include that the exposure time interval of the first product is a preset time interval; Presentation layer 3 is specifically used for: Get the ranking results of the scores of each candidate product from large to small; If the exposure time interval of the first product corresponding to the target quantity score ranked at the top in the sorting result is not the preset time interval, the score of the first product will be removed from the sorting result, and the product corresponding to the target quantity score ranked at the top will be reselected as the product recommended for exposure to the target user this time; the target quantity is equal to the displayable quantity of the product.
[0057] In one embodiment, the limited exposure conditions of different products include that the maximum exposure duration of the second product is a preset duration; Presentation layer 3 is specifically used for: Get the ranking results of the scores of each candidate product from large to small; If the exposure time of the second product corresponding to the target number score ranked at the top in the sorting result reaches the preset time, the score of the second product will be removed from the sorting result, and the product corresponding to the target number score ranked at the top will be reselected as the product recommended for exposure to the target user this time; the target number is equal to the displayable number of products.
[0058] In one embodiment, limiting exposure conditions for different products includes limiting continuous exposure of a third product within a preset period; Presentation layer 3 is specifically used for: Get the ranking results of the scores of each candidate product from large to small; If the third product corresponding to the target number score ranked high in the sorting result is continuously exposed within the preset period, the score of the third product will be removed from the sorting result, and the product corresponding to the target number score ranked high will be reselected as the product recommended to the target user this time; the target number is equal to the displayable number of products.
[0059] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the exposure recommendation method, which includes: responding to the access request sent by the target user, collecting the original basic data and processing the original basic data to obtain the target basic data; the original basic data includes the original user data and the original product data; recalling the product based on the target basic data to obtain the candidate product set; for each candidate product in the candidate product set, using the conversion rate model to estimate the first conversion probability of the candidate product for the target user, and using the exposure value model to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time; weighting the first conversion probability and the second conversion probability to obtain the score of the candidate product; based on the score of each candidate product, combined with the exposure position and the business control strategy, determine the product recommended for exposure to the target user this time and display it.
[0060] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0061] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the exposure recommendation method provided by the above-mentioned method embodiments, the method including: responding to an access request sent by a target user, collecting original basic data and processing the original basic data to obtain target basic data; the original basic data includes original user data and original product data; recalling products based on the target basic data to obtain a set of candidate products; for each candidate product in the candidate product set, using a conversion rate model to estimate the first conversion probability of the candidate product for the target user, and using an exposure value model to estimate the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time; weighted fusion of the first conversion probability and the second conversion probability to obtain the score of the candidate product; based on the score of each candidate product, combined with the exposure position and business control strategy, determine the product recommended for exposure to the target user this time and display it.
[0062] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the exposure recommendation method provided by the above-mentioned method embodiments, the method comprising: responding to an access request sent by a target user, collecting original basic data and processing the original basic data to obtain target basic data; the original basic data comprises original user data and original product data; recalling products based on the target basic data to obtain a set of candidate products; for each candidate product in the candidate product set, using a conversion rate model to estimate a first conversion probability of the candidate product for the target user, and using an exposure value model to estimate a second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time; weightedly fusing the first conversion probability and the second conversion probability to obtain a score for the candidate product; based on the score of each candidate product, combined with the exposure position and the business control strategy, determining and displaying the product recommended for exposure to the target user this time.
[0063] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0064] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. An exposure recommendation method, characterized in that: include: In response to the access request sent by the target user, the original basic data is collected and processed to obtain the target basic data; The original basic data includes original user data and original product data; Perform product recall based on the target basic data to obtain a candidate product set; For each candidate product in the candidate product set, a first conversion probability of the candidate product for the target user is estimated using a conversion rate model, and a second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is estimated using an exposure value model; Performing a weighted fusion of the first conversion probability and the second conversion probability to obtain a score of the candidate product; Based on the scores of the candidate products, combined with the exposure position and the business control strategy, the recommended products to be exposed to the target users are determined and displayed.
2. The exposure recommendation method according to claim 1, characterized in that: The exposure value model is trained by the following steps: Input the historical exposure information of different products on the product side for different users, the conversion rates after different historical exposure times, and the historical exposure sequences of different users on the user side who exposed different products at different times into the recurrent neural network, and predict the conversion probability after different users are exposed to different products this time; the historical exposure information includes at least one of the historical exposure times, historical exposure duration, historical exposure interval, and other products that were exposed at the same time in history; The recurrent neural network is trained based on each of the predicted conversion probabilities and the corresponding actual conversion results until the model converges to obtain the exposure value model.
3. The exposure recommendation method according to claim 2, characterized in that: The estimating the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time by using the exposure value model includes: The historical exposure information of the candidate product on the product side for the target user, the conversion rate after different historical exposure times, and the historical exposure sequence of the target user on the user side when exposing the candidate product at different times are input into the exposure value model, and the second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is output.
4. The exposure recommendation method according to claim 1, characterized in that: Based on the scores of the candidate products, combined with the exposure position and the business control strategy, determining the recommended product to be exposed to the target user and displaying it, includes: Determining exposure restrictions for different products based on the business control strategy; Determine the displayable quantity of the product based on the exposure position; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products; The product recommended for exposure to the target user this time is displayed at the exposure position.
5. The exposure recommendation method according to claim 4, characterized in that: The restricted exposure conditions of different products include that the exposure time interval of the first product is a preset time interval; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the exposure time interval of the first product corresponding to the score of the target quantity ranked higher in the sorting result is not the preset time interval, the score of the first product is removed from the sorting result, and the product corresponding to the score of the target quantity ranked higher is reselected as the product recommended for exposure to the target user this time; the target quantity is equal to the displayable quantity of the product.
6. The exposure recommendation method according to claim 4, characterized in that: The limited exposure conditions of different products include that the maximum exposure time of the second product is a preset time; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the exposure time of the second product corresponding to the score of the target number ranked at the top in the sorting result reaches the preset time, the score of the second product will be removed from the sorting result, and the product corresponding to the score of the target number ranked at the top will be reselected as the product recommended for exposure to the target user this time; the target number is equal to the displayable number of the product.
7. The exposure recommendation method according to claim 4, characterized in that: The restricted exposure conditions of the different products include limiting the continuous exposure of the third product within a preset period; Under the constraints of the limited exposure of the different products and the displayable quantity of the products, the products recommended for exposure to the target user are selected from the ranking results of the scores of the candidate products, including: Obtain a ranking result of the scores of each candidate product from largest to smallest; If the third product corresponding to the target number score ranked high in the sorting result is continuously exposed within the preset period, the score of the third product will be removed from the sorting result, and the product corresponding to the target number score ranked high will be reselected as the product recommended for exposure to the target user this time; the target number is equal to the displayable number of the product.
8. An exposure recommendation system, characterized in that: include: The data layer is used to respond to the access request sent by the target user, collect the original basic data and process the original basic data to obtain the target basic data; the original basic data includes original user data and original product data; The model layer is used for: Perform product recall based on the target basic data to obtain a candidate product set; For each candidate product in the candidate product set, a first conversion probability of the candidate product for the target user is estimated using a conversion rate model, and a second conversion probability of the candidate product for the target user after the candidate product is exposed to the target user this time is estimated using an exposure value model; Performing a weighted fusion of the first conversion probability and the second conversion probability to obtain a score of the candidate product; The display layer is used to determine and display the recommended products to be exposed to the target user based on the scores of the candidate products, combined with the exposure position and business control strategy.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the exposure recommendation method according to any one of claims 1 to 7 is implemented.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the exposure recommendation method according to any one of claims 1 to 7 is implemented.