Product recommendation strategy determination method and device, electronic equipment and storage medium

By obtaining the conversion probability and expected returns of the products to be recommended, and optimizing the product recommendation strategy matrix, the problem of uneven resource allocation under the manual recommendation product method is solved, and the accuracy and efficiency of the product recommendation strategy are improved.

CN119048202AInactive Publication Date: 2024-11-29中国联合网络通信有限公司广东省分公司
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411176505.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The manual product recommendation method in the prior art makes it difficult to promote service products with low priority, unable to effectively recommend and promote product, and unbalanced resource allocation.

Method used

By obtaining the product recommendation strategy matrix, conversion probability and expected returns of products to be recommended during the current recommendation cycle, determining the recommendation strategy score function and constraints, optimizing product recommendation strategy based on this information, and ensuring reasonable allocation of resources.

Benefits of technology

The target product recommendation strategy is achieved to achieve accurate and convenient determination of the recommended product, improve the efficiency and accuracy of the recommendation strategy, and ensure the improvement of user perception and company profits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048202A_ABST
    Figure CN119048202A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a product recommendation strategy determination method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a current product recommendation strategy matrix and a current recommendation demand of a to-be-recommended product in a current recommendation period, and a current conversion probability and a current expected income of a to-be-recommended user for each to-be-recommended product; based on the current product recommendation strategy matrix, the current conversion probability and the current expected income, determining a recommendation strategy score function corresponding to each to-be-recommended product, and analyzing the current recommendation demand to determine a constraint condition corresponding to the recommendation strategy score function; determining a current recommendation strategy score corresponding to the current product recommendation strategy matrix based on the current product recommendation strategy matrix, a recommendation strategy score function and a constraint condition, and determining a target product recommendation strategy corresponding to each to-be-recommended product based on the current recommendation strategy score and a preset convergence condition, and the efficiency and accuracy of determining the target product recommendation strategy are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of resource allocation, and in particular, to a method, device, electronic device and storage medium for determining a product recommendation strategy. Background Art

[0002] With the development of information technology, users' demands for network, voice and information services provided by operators are becoming more and more diverse. To this end, operators have launched a variety of service products to cover user needs, and promoted service products to users through various recommendation activities to improve user perception and company profits. Since recommendation activities consume a lot of recommendation resources and there are limited recommendation opportunities for users, it is particularly important to make accurate recommendations to users.

[0003] Currently, manual product recommendation is usually adopted to recommend service products to users based on manual business experience. Operators can determine the priority of service products based on business experience and recommend service products to users in descending order according to the priority of service products.

[0004] However, this method of manually recommending products allows service products with higher priority to obtain more high-quality user resources. Conversely, service products with lower priority will be difficult to promote due to fewer user resources, resulting in the inability to make effective product recommendations and promotions. Summary of the invention

[0005] The embodiments of the present invention provide a product recommendation strategy determination method, device, electronic device and storage medium to accurately and conveniently determine the target product recommendation strategy corresponding to all products to be recommended in the current recommendation cycle, thereby improving the efficiency and accuracy of determining the target product recommendation strategy.

[0006] In a first aspect, an embodiment of the present invention provides a method for determining a product recommendation strategy, comprising:

[0007] Obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected return of the recommended user for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected return is predicted based on the target product expected return prediction model;

[0008] Based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, determine the recommendation strategy score function corresponding to each to-be-recommended product, and determine the constraint condition corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each to-be-recommended product;

[0009] Based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, the current recommendation strategy score corresponding to the current product recommendation strategy matrix is ​​determined, and based on the current recommendation strategy score and the preset convergence condition, the target product recommendation strategy corresponding to each product to be recommended is determined.

[0010] In a second aspect, an embodiment of the present invention further provides a device for determining a product recommendation strategy, the device comprising:

[0011] An information acquisition module is used to obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected return of the recommended user for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected return is predicted based on the target product expected return prediction model;

[0012] A recommendation strategy score function determination module is used to determine the recommendation strategy score function corresponding to each to-be-recommended product based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, and determine the constraint conditions corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each to-be-recommended product;

[0013] The target product recommendation strategy determination module is used to determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, and to determine the target product recommendation strategy corresponding to each product to be recommended based on the current recommendation strategy score and the preset convergence condition.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0015] one or more processors;

[0016] A memory for storing one or more programs;

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement the product recommendation strategy determination method provided by any embodiment of the present invention.

[0018] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for determining a product recommendation strategy as provided in any embodiment of the present invention.

[0019] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements a method for determining a product recommendation strategy as provided in any embodiment of the present invention.

[0020] The technical solution of the embodiment of the present invention obtains the current product recommendation strategy matrix and the current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and the current expected benefit of the recommended users for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected benefit is predicted based on the target product expected benefit prediction model; based on the current product recommendation strategy matrix, the current conversion probability and the current expected benefit, the recommendation strategy score function corresponding to each product to be recommended is determined, and by analyzing the current recommendation demand of each product to be recommended, the constraint condition corresponding to the recommendation strategy score function is determined; based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint condition, the current recommendation strategy score corresponding to the current product recommendation strategy matrix is ​​determined, and based on the current recommendation strategy score and the preset convergence condition, the target product recommendation strategy is uniformly determined for all products to be recommended, so as to accurately and conveniently determine the target product recommendation strategy corresponding to all products to be recommended in the current recommendation cycle, and improve the efficiency and accuracy of determining the target product recommendation strategy, so that the products to be recommended can be effectively recommended to the users to be recommended according to the target product recommendation strategy, so as to improve user perception and company benefits.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flow chart of a method for determining a product recommendation strategy provided by Embodiment 1 of the present invention;

[0024] Figure 2 is a flow chart of a method for determining a product recommendation strategy provided by Embodiment 2 of the present invention;

[0025] Figure 3It is a structural diagram of a device for determining a product recommendation strategy provided by Embodiment 3 of the present invention;

[0026] Figure 4 It is a structural schematic diagram of an electronic device that implements the product recommendation strategy determination method of an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment 1

[0030] Figure 1 A flowchart of a method for determining a product recommendation strategy is provided for the first embodiment of the present invention. This embodiment is applicable to the case of uniformly determining the target product recommendation strategy corresponding to all products to be recommended. The method can be executed by a product recommendation strategy determination device, which can be implemented in the form of hardware and / or software, and can be configured in an electronic device. Figure 1 As shown, the method includes:

[0031] S110, obtaining the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected benefit of the recommended user for each product to be recommended.

[0032] Among them, the current conversion probability is predicted based on the target product conversion probability prediction model. The current expected return is predicted based on the target product expected return prediction model. The products to be recommended and the users to be recommended will change. For example, if a new product is launched, the product can be added to the products to be recommended. If a product is ineffective, the product is removed from the shelves and no longer recommended. Newly registered users can be added to the users to be recommended. The current recommendation cycle can refer to the time period for recommending existing products to be recommended to existing users to be recommended. There are new products and users during the current recommendation cycle, but the new products and users are not considered in the current recommendation cycle. The products to be recommended and the users to be recommended can be updated at the beginning of the next recommendation cycle, and the target product recommendation strategy for the next recommendation cycle can be formulated for the updated products to be recommended and the users to be recommended. The products to be recommended can refer to service products to be recommended to users. For example, the products to be recommended can be but are not limited to products on sale. The current product recommendation strategy matrix can be used to characterize the current product recommendation strategy. The current product recommendation strategy can be a strategy for which user to recommend which product. Each element in the current product recommendation strategy matrix can indicate whether a product is recommended to a certain user. The value of each element in the current product recommendation strategy matrix is ​​zero or one. The element value is one, which may indicate that a certain product is recommended to a certain user. The element value is zero, which may indicate that a certain product is not recommended to a certain user. The current recommendation demand may refer to the recommendation demand of the product to be recommended itself, that is, the indicator that the product to be recommended needs to achieve when recommending the product to be recommended to the user. For example, the current recommendation demand may include but is not limited to the recommendation quantity interval limit, the current conversion probability limit and the current expected return limit of the product to be recommended. The current conversion probability may refer to the probability that the user uses the product to be recommended. The current conversion probability can also be understood as the probability that a non-product user is converted into a product user. The current expected return may refer to the amount of money that the user consumes for the recommended product. The target product conversion probability prediction model can use the product to be recommended and the user to be recommended in the current recommendation cycle to predict the current conversion probability of each user to be recommended for each product to be recommended. For example, the product features of the product to be recommended and the user features of the user to be recommended can be used as the input of the target product conversion probability prediction model. The target product expected return prediction model can use the product to be recommended and the user to be recommended in the current recommendation cycle to predict the current expected return of each user to be recommended for each product to be recommended. For example, the product features of the product to be recommended and the user features of the user to be recommended may be used as inputs of the target product expected revenue prediction model.

[0033] Specifically, the current product recommendation strategy matrix of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected benefits of the recommended users for each product to be recommended can be obtained from the pre-stored information, and the current recommendation requirements of the products to be recommended can be determined based on the performance indicators that each product to be recommended needs to achieve. For example, in order to ensure that the performance indicators can be achieved, such as converting 20 non-product users into product users, the products to be recommended need to be recommended to at least 40 users to be recommended, and the current conversion probability of each recommended user cannot be less than 50%.

[0034] On the basis of the above technical solution, "obtaining the current product recommendation strategy matrix of products to be recommended in the current recommendation cycle" may include: obtaining the number of products to be recommended and the number of users to be recommended in the current recommendation cycle, and based on the number of products and the number of users, constructing the current product recommendation strategy matrix of products to be recommended in the current recommendation cycle; wherein the elements in the current product recommendation strategy matrix are used to represent the current recommendation status of users to be recommended and products to be recommended.

[0035] The number of products may refer to the total number of products to be recommended. The number of users may refer to the total number of users to be recommended. The current recommendation status may refer to whether the product to be recommended is recommended to the user to be recommended.

[0036] Specifically, the number of products to be recommended in the current recommendation cycle, m, and the number of users to be recommended, n, are obtained, and the number of products is used as the total number of columns in the current product recommendation strategy matrix, and the number of users is used as the total number of rows in the current product recommendation strategy matrix to construct the current product recommendation strategy matrix for the products to be recommended in the current recommendation cycle. For example, the current product recommendation strategy matrix can be expressed as Among them, x i,j ∈{0,1}. i,j =0, which means that the jth product is not recommended to the i-th user. i,j =1, which can be expressed as recommending the jth product to the i-th user. The advantage of constructing the current product recommendation strategy matrix is ​​that it can consider all users and all products at the same time, avoiding the situation where some products are recommended to users first due to priorities, resulting in other products not being effectively recommended. That is, a reasonable recommendation strategy is determined in an overall form, further improving the accuracy of the target product recommendation strategy.

[0037] On the basis of the above technical solution, "obtaining the current conversion probability and current expected revenue of the user to be recommended for each product to be recommended in the current recommendation cycle" may include: obtaining the current user information of the user to be recommended and the current product information of the product to be recommended in the current recommendation cycle; inputting the current user information and the current product information into the target product conversion probability prediction model for probability prediction, and determining the current conversion probability of the user to be recommended for the product to be recommended in the current recommendation cycle based on the output of the target product conversion probability prediction model; the target product conversion probability prediction model is pre-trained based on a preset classification model; inputting the current user information and the current product information into the target product expected revenue prediction model for revenue prediction, and determining the current expected revenue of the user to be recommended for the product to be recommended in the current recommendation cycle based on the output of the target product expected revenue prediction model; the target product expected revenue prediction model is pre-trained based on a preset regression model.

[0038] Among them, the current user information may include but is not limited to user basic attributes, user usage behavior, user product ordering records, and user account information. The current product information may include but is not limited to product basic information and historical order summary statistics. The preset classification model may refer to a model containing any classification algorithm such as random forest or decision tree that can output prediction probability. The preset regression model may refer to a model containing numerical regression algorithms such as logistic regression. The preset classification model and the preset regression model are used by machine learning to obtain the target product conversion probability prediction model and the target product expected revenue prediction model. The traditional scoring model is only established for a single scene or a single product, and there are problems such as difficult model operation and maintenance, poor effect of small scenes, and cold start of new scenes. For example, when the number of scenes and targets increases, hundreds of independent models greatly increase system resource overhead and labor costs, which will lead to difficult model operation and maintenance. Due to insufficient training data, the model accuracy for small scenes is insufficient, which will lead to poor effect of small scenes. For new scenes, due to the lack of training samples, new scenes will be cold started. The technical solution of the embodiment of the present invention can build a target product conversion probability prediction model and a target product expected profit prediction model through recommendation records of each scenario, integrating product, scenario, and user characteristics, so as to realize one model to predict all scenarios and targets.

[0039] Specifically, the current user information of the user to be recommended and the current product information of the product to be recommended in the current recommendation cycle are obtained, and the features of the current user information and the current product information are extracted to obtain the user features and the product features. The user features and the product features are input into the target product conversion probability prediction model for probability prediction to obtain the current conversion probability of each user for each product in the current recommendation cycle, that is, (user i ,product j ) corresponds to the current conversion probability p ijInput the user characteristics and product characteristics into the target product expected revenue prediction model to predict the revenue, and obtain the current expected revenue of each user for each product in the current recommendation cycle, that is, (user i ,product j ) corresponds to the current expected return m ij Table 1 gives an example of the output results of a model, as follows:

[0040] Table 1 An example of a model output result

[0041] User ID Product ID Current conversion probability Current expected return Recommended expected benefits Sorting u1 p1 0.53 3.13 1.6589 1 u1 p3 0.34 4.0 1.36 2 u1 p2 0.02 10 0.2 3

[0042] Among them, the current conversion probability p ij With the current expected return m ij Multiply them to get the expected benefit e of the recommended product for the recommended user. ij Among them, e ij =p ij *m ij After determining the expected benefit of the recommendation, the method can also use the expected benefit of the recommendation to perform sorting to obtain the recommendation with the highest expected benefit.

[0043] Exemplarily, the process of using machine learning to perform on a preset classification model and a preset regression model to obtain a target product conversion probability prediction model and a target product expected revenue prediction model includes: obtaining historical product information of the product to be recommended, historical user information of the user to be recommended, and historical recommendation record data within the historical recommendation cycle; extracting features from the historical product information to obtain historical product features, and extracting features from the historical user information to obtain historical user features, and constructing a feature wide table based on the fusion of historical product features and historical user features. Extract historical recommendation record data as sample labels for model training, and associate the sample labels with the feature wide table to form sample data. Using the training data set, train the user-product recommendation conversion probability prediction model based on the preset classification model, and train the user-product expected revenue model based on the preset regression model. Use the test data set to calculate the performance indicators of the above models, and select the model with the highest accuracy as the target product conversion probability prediction model and the target product expected revenue prediction model.

[0044] Among them, historical user information may include user basic attributes, user usage behavior, user product ordering records and user account information. Historical product information may include product basic information and historical order summary statistics. Historical recommendation record data may include whether to order (such as whether the user orders the product after recommending it to the user), the amount of money billed after ordering (such as the amount of money billed in month t+1 after ordering) and whether to cancel the order after ordering. Sample data includes training data set and test data set.

[0045] Specifically, for the preset classification model, the area (Area Under Curve, AUC) enclosed by the coordinate axis under the characteristic curve (receiver operating characteristic curve, ROC curve) can be selected as the model effect evaluation indicator. The calculation process of AUC includes: sorting the prediction scores output by the model, calculating the true positive rate (True Positive Rate, TPR) and false positive rate (False Positive Rate, FPR) of each score point, drawing the ROC curve, and calculating the area under the curve. The value of AUC is usually between 0.5 and 1. The higher the AUC, the better the model classification effect, that is, the more accurate the prediction. For the preset regression model, the mean absolute error (Mean Absolute Deviation, MAE) can be used to measure the model effect, such as the absolute average error of the difference between the predicted income value-added and the actual income change. The smaller the absolute average error, the better the prediction effect of the model.

[0046] Exemplarily, before constructing the feature wide table, the data can be preprocessed, including: missing value filling, outlier processing, data type conversion, and data merging. Among them, missing value filling includes using the mode filling for continuous features and using "unknown" filling for discrete features. Outlier processing includes calculating and correcting outliers for continuous features using the box plot calculation method. Data type conversion includes using heat encoding to convert discrete features into 0-1 representation. Finally, all feature data are merged to construct a feature wide table as the model input feature.

[0047] S120. Based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, determine the recommendation strategy score function corresponding to each product to be recommended, and determine the constraint conditions corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each product to be recommended.

[0048] The recommendation strategy score function may refer to a function for calculating the overall score of the current product recommendation strategy matrix. The constraint condition may refer to a constraint condition for the solution of the recommendation strategy score function.

[0049] Specifically, the element value x corresponding to each element in the current product recommendation strategy matrix i,j , current conversion probability p ij and the current expected return m ij Multiply to get the score x corresponding to each element i,j *p ij *m ij , and add the scores corresponding to all elements in the current product recommendation strategy matrix to construct the recommendation strategy score function corresponding to each product to be recommended By analyzing the current recommendation demand for each product to be recommended, determine the constraints corresponding to the recommendation strategy score function. For example, if the current recommendation demand is that each user can only be recommended one product at most, the constraint can be

[0050] For example, targeted scoring functions and constraints can be set according to different business requirements. For example, if cost is considered, the scoring function can be Among them, c ij It can represent the recommendation cost required to recommend a product to a user.

[0051] On the basis of the above technical solution, "determining the recommendation strategy score function corresponding to each product to be recommended based on the current product recommendation strategy matrix, the current conversion probability and the current expected benefit" may include: multiplying the current conversion probability by the current expected benefit to obtain the expected benefit of the recommendation of the recommended user for the product to be recommended; multiplying each element in the current product recommendation strategy matrix by the expected benefit of the recommendation corresponding to each element to obtain the recommendation numerator function corresponding to each element; adding the recommendation numerator functions corresponding to each element to obtain the recommendation strategy score function corresponding to each product to be recommended.

[0052] The expected benefit of recommendation may refer to the benefit increased by recommending the product to the user. The recommendation score subfunction may refer to the score subfunction corresponding to each element in the current product recommendation strategy matrix.

[0053] Specifically, the current conversion probability p ij With the current expected return m ij Multiply them to get the expected benefit e of the recommended product for the recommended user. ij Among them, e ij =p ij *m ij . Each element x in the current product recommendation strategy matrix i,j The recommended expected benefit e corresponding to each element ij Multiply them to get the recommended numerator function x for each element i,j *e ij , and add the recommended subfunctions corresponding to each element to obtain the recommendation strategy score function corresponding to each product to be recommended

[0054] Based on the above technical solution, the constraints include at least one of the following: the sum of the element values ​​of each row of the current product recommendation strategy matrix is ​​one, and each element is zero or one; the sum of the element values ​​of each column of the current product recommendation strategy matrix is ​​within the preset recommendation quantity range in the current recommendation demand; the elements in the current product recommendation strategy matrix whose current conversion probability is less than or equal to the preset conversion probability in the current recommendation demand are set to zero; the elements in the current product recommendation strategy matrix whose current expected return is less than or equal to the preset expected return in the current recommendation demand are set to zero.

[0055] The preset recommendation quantity interval may refer to the recommendation quantity interval determined based on the current recommendation demand. The preset conversion probability may refer to the conversion probability threshold determined based on the current recommendation demand. The preset expected return may refer to the expected return threshold determined based on the current recommendation demand. The above constraints may be used to narrow the product recommendation range, i.e., the recommended user range.

[0056] Specifically, if the current recommendation requirements include: the upper limit of the number of recommended users and the upper limit of the number of recommended products N_MAX j and lower limit N_MIN j , the lower limit of the current conversion probability score P_MIN j And the lower limit of the current expected return score M_MIN j , where j is product j. The constraints corresponding to the recommendation strategy score function can be determined by analyzing the current recommendation demand of each product to be recommended. For the upper limit of the number of recommendations for a user, the corresponding constraint is that the sum of the element values ​​of each row of the current product recommendation strategy matrix is ​​one, and each element is zero or one. For the upper limit of the number of product recommendations N_MAX j and lower limit N_MIN j The corresponding constraint condition is that the sum of the element values ​​of each column in the current product recommendation strategy matrix is ​​within the preset recommendation quantity range [N_MIN j , N_MAX j ] The lower limit of the score P_MIN for the current conversion probability j , and the corresponding constraint condition is that the current conversion probability in the current product recommendation strategy matrix is ​​less than or equal to the preset conversion probability P_MIN in the current recommendation demand j The elements of are set to zero. The lower limit of the score for the current expected return is M_MIN j The corresponding constraint condition is that the current expected return in the current product recommendation strategy matrix is ​​less than or equal to the preset expected return M_MIN in the current recommendation demand. j Set the elements of to zero.

[0057] It should be noted that due to different business scenario requirements, different scenarios have different requirements for the number of recommended customer groups and user quality. When outputting product recommendation strategies, it is necessary to determine the requirements of each product based on the current recommendation requirements, so as to determine the constraints corresponding to the recommendation strategy score function. Table 2 gives an example of the current recommendation requirements of a product to be recommended, as follows:

[0058] Table 2 An example of the current recommendation demand for a product to be recommended

[0059]

[0060]

[0061] Among them, product p1 requires the number of users to be between 120 and 300, the preset conversion probability to be no less than 30%, and the preset expected income to be no less than 10 yuan.

[0062] For example, the constraint can be expressed as

[0063] S130. Based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraints, determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix, and based on the current recommendation strategy score and the preset convergence condition, determine the target product recommendation strategy corresponding to each product to be recommended.

[0064] Among them, the current recommendation strategy score may refer to the score of the current product recommendation strategy matrix in the recommendation strategy score function. The current recommendation strategy score may be used to represent the score corresponding to the current product recommendation strategy corresponding to the current product recommendation strategy matrix. The preset convergence condition may refer to a preset iteration end condition. For example, the preset convergence condition may be, but is not limited to, the current total number of iterations reaching a preset iteration number threshold or the rate of change between the current recommendation strategy score and the recommendation strategy score corresponding to the previous preset number of iterations being within a preset change range. The target product recommendation strategy may refer to the most effective product recommendation strategy in the current recommendation cycle. For example, the target product recommendation strategy corresponds to the highest target recommendation strategy score.

[0065] Specifically, the element values ​​of all elements in the current product recommendation strategy matrix are substituted into the recommendation strategy score function to obtain the current function value corresponding to the current product recommendation strategy matrix, that is, the current recommendation strategy score. Check whether the current recommendation strategy score meets the preset convergence condition. If it does, the current product recommendation strategy matrix is ​​used as the target product recommendation strategy matrix, and the target product recommendation strategy corresponding to each product to be recommended is determined based on the target product recommendation strategy matrix. If it does not meet the condition, the element values ​​in the current product recommendation strategy matrix are iteratively updated according to the constraint conditions to obtain the iteratively updated current product recommendation strategy matrix, and the current recommendation strategy score is continued to be calculated based on the iteratively updated current product recommendation strategy matrix until it is detected that the current recommendation strategy score meets the preset convergence condition, and then the iteration is stopped.

[0066] The technical solution of the embodiment of the present invention obtains the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected profit of the recommended users for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected profit is predicted based on the target product expected profit prediction model; based on the current product recommendation strategy matrix, the current conversion probability and the current expected profit, the recommendation strategy score function corresponding to each product to be recommended is determined, and by analyzing the current recommendation demand of each product to be recommended, the constraint conditions corresponding to the recommendation strategy score function are determined; based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, the current recommendation strategy score corresponding to the current product recommendation strategy matrix is ​​determined, and based on the current recommendation strategy score and the preset convergence conditions, the target product recommendation strategy is uniformly determined for all products to be recommended, so as to accurately and conveniently determine the target product recommendation strategy corresponding to all products to be recommended in the current recommendation cycle, and improve the efficiency and accuracy of determining the target product recommendation strategy, so that the products to be recommended can be effectively recommended to the users to be recommended according to the target product recommendation strategy, so as to improve user perception and company benefits.

[0067] It should be noted that the recommendation in the embodiment of the present invention can be understood as marketing.

[0068] Embodiment 2

[0069] Figure 2 This is a flowchart of a method for determining a product recommendation strategy provided in the second embodiment of the present invention. Based on the above embodiments, this embodiment describes in detail the process of determining the target product recommendation strategy corresponding to each product to be recommended. The explanations of the terms that are the same or corresponding to the above embodiments are not repeated here. Figure 2 As shown, the method includes:

[0070] S210: Obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected benefit of the recommended user for each product to be recommended.

[0071] The current conversion probability is predicted based on the target product conversion probability prediction model, and the current expected return is predicted based on the target product expected return prediction model.

[0072] S220. Based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, determine the recommendation strategy score function corresponding to each product to be recommended, and determine the constraint conditions corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each product to be recommended.

[0073] S230. Iteratively adjust the current product recommendation strategy matrix based on preset adjustment rules and constraints to obtain the current product recommendation strategy matrix after element values ​​are adjusted.

[0074] The constraint condition can be understood as the constraint condition for updating and iterating the current product recommendation strategy matrix. The current product recommendation strategy matrix can be understood as the solution of the recommendation strategy score function. The preset adjustment rule can refer to the preset adjustment rule of the element value in the current product recommendation strategy matrix. For example, the preset adjustment rule can be, but is not limited to, a genetic algorithm.

[0075] Specifically, based on the genetic algorithm, the element value of an element that meets the selection conditions in the current product recommendation strategy matrix is ​​mutated (the element value is changed from zero to one or the element value is changed from one to zero), or the element values ​​of two elements that meet the selection conditions in the current product recommendation strategy matrix are crossed (the element values ​​of two elements are exchanged) to obtain the current product recommendation strategy matrix after the element values ​​are adjusted.

[0076] On the basis of the above technical solution, "iteratively adjusting the current product recommendation strategy matrix based on preset adjustment rules and constraints to obtain the current product recommendation strategy matrix after element values ​​are adjusted" may include: dividing the current product recommendation strategy matrix based on a preset number of user subgroups to obtain a preset number of candidate user subgroups; determining the elements to be adjusted in the target user subgroup based on a preset crossover probability, a preset mutation probability and a preset selection probability corresponding to each element in each candidate user subgroup; the preset selection probability is determined based on a numerator function of the recommendation score corresponding to each element in each candidate user subgroup; setting the element values ​​of the elements to be adjusted to zero or one to obtain the current product recommendation strategy matrix after element values ​​are adjusted.

[0077] The preset number of user subgroups may refer to the number of subgroups preset for dividing the users to be recommended. For example, the preset number of user subgroups may be k. The user subgroups to be selected may refer to the user subgroups obtained after the users to be recommended are divided according to the preset number of user subgroups. The user subgroups to be selected may be represented by X i . The advantage of this setting is that the overall product recommendation strategy score can be converted into the sum of the scores of each candidate user subgroup. The preset crossover probability may refer to the probability of a crossover operation in a preset element value adjustment process. The preset mutation probability may refer to the probability of a mutation operation in a preset element value adjustment process. The preset selection probability may be a selection probability determined based on the recommended subfunction corresponding to each element in each candidate user subgroup. The element to be adjusted may refer to the element whose element value needs to be adjusted during this iteration.

[0078] Specifically, the number of user subgroups is preset to k, and the recommended users are randomly decomposed into k different selected user subgroups. The overall product recommendation strategy score is converted into the sum of the scores of each selected user subgroup, that is: For each candidate user subgroup, set the population size of the candidate user subgroup to q, and initialize the population P of the candidate user subgroup i i = {P i,1 ,…,P i,q}, replace the corresponding subgroup in the current product recommendation strategy matrix X for each individual in the population, and calculate the fitness F(P ij )=F(X1,…,X i-1 ,P ij ,X i+1 ,X k ). The genetic operation is performed on the population in a loop until the maximum number of loops or other termination conditions are reached. When the loop ends, the current product recommendation strategy matrix obtained by the loop is determined as the optimal solution of the product recommendation strategy, and the target product recommendation strategy corresponding to each product to be recommended is determined based on the optimal solution.

[0079] Exemplarily, performing genetic operations on the population in a loop until a maximum number of loops or other termination conditions are reached may include: for each population of the user subgroup to be selected, performing genetic operations using a preset crossover probability, a preset mutation probability, and a preset selection probability, and performing genetic operations on the population P of the user subgroup to be selected. i Update and obtain the offspring population O i Among them, the probability of each individual being selected is preset, that is, the preset selection probability, which is Randomly select two individuals and crossover points, generate offspring, and then perform mutation with a preset probability p. m Modify the value of a certain position in the offspring, that is, the element value. i, calculate the fitness F(O ij ) = F(X1, …, X i-1 , O ij , X i+1 , X k ), and perform detection. If it is detected that F(O ij ) < F(X) and X new = {X1, …, X i-1 , O ij , X i+1 , X k} satisfies the constraint conditions, then update X = X new , F(X) = F(O ij ). Generate the population P i of the next generation of candidate user subgroups from O i and P i , and continue to loop until the maximum number of loops or other termination conditions are reached.

[0080] It should be noted that the candidate user subgroups can be obtained through different partitioning methods. For example, the candidate user subgroups can be obtained by partitioning users according to dimensions such as user attributes.

[0081] S240. Based on the current product recommendation strategy matrix and the recommendation strategy score function after adjusting the element values, determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix after adjusting the element values.

[0082] Specifically, substitute all the element values in the current product recommendation strategy matrix after adjusting the element values into the recommendation strategy score function to obtain the current recommendation strategy score corresponding to the current product recommendation strategy matrix after adjusting the element values.

[0083] S250. When the current recommendation strategy score meets the preset convergence condition, determine the target product recommendation strategies corresponding to each product to be recommended based on the element values of each element in the current product recommendation strategy matrix.

[0084] Specifically, when the current recommendation strategy score meets the preset convergence condition, stop the iteration of the current product recommendation strategy matrix, and the finally obtained current product recommendation strategy matrix can be used as the optimal solution, and determine the target product recommendation strategies corresponding to each product to be recommended based on the element values of each element in the finally obtained current product recommendation strategy matrix.

[0085] The technical solution of the embodiment of the present invention iteratively adjusts the current product recommendation strategy matrix based on preset adjustment rules and constraints to obtain the current product recommendation strategy matrix after element value adjustment, and determines the current recommendation strategy score corresponding to the current product recommendation strategy matrix after element value adjustment after each iteration based on the current product recommendation strategy matrix after element value adjustment and the recommendation strategy score function, thereby performing conditional judgment based on the current recommendation strategy score, and when the current recommendation strategy score meets the preset convergence condition, the current product recommendation strategy matrix corresponding to the current recommendation strategy score is used as the optimal solution of the recommendation strategy score function, and the target product recommendation strategy corresponding to each to-be-recommended product is determined based on the element value of each element in the current product recommendation strategy matrix, thereby taking all to-be-recommended products into consideration at the same time, and effectively recommending the to-be-recommended products to the to-be-recommended users according to the target product recommendation strategy corresponding to the optimal solution, so as to improve user perception and company benefits, and further improve the accuracy of determining the target product recommendation strategy.

[0086] It should be noted that when the scale of user resources and product resources is large (such as a user scale of tens of millions and a product scale of thousands), traditional optimization algorithms cannot converge or fall into local optimal values. The technical solution of the embodiment of the present invention can use a collaborative co-evolutionary algorithm to decompose large-scale optimization problems into several groups of problems according to variables, optimize them in groups, and cooperate and collaborate between the groups to jointly complete the optimization of the entire problem, thereby obtaining an accurate optimal solution.

[0087] The following is an embodiment of a product recommendation strategy determination device provided in an embodiment of the present invention. The device and the product recommendation strategy determination methods of the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the product recommendation strategy determination device, please refer to the embodiment of the above-mentioned product recommendation strategy determination method.

[0088] Embodiment 3

[0089] Figure 3 This is a schematic diagram of the structure of a device for determining a product recommendation strategy provided by Embodiment 3 of the present invention. Figure 3 As shown, the device includes: an information acquisition module 310, a recommendation strategy score function determination module 320 and a target product recommendation strategy determination module 330.

[0090] Among them, the information acquisition module 310 is used to obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected profit of the recommended users for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected profit is predicted based on the target product expected profit prediction model; the recommendation strategy score function determination module 320 is used to determine the recommendation strategy score function corresponding to each product to be recommended based on the current product recommendation strategy matrix, the current conversion probability and the current expected profit, and determine the constraint conditions corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each product to be recommended; the target product recommendation strategy determination module 330 is used to determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, and determine the target product recommendation strategy corresponding to each product to be recommended based on the current recommendation strategy score and the preset convergence conditions.

[0091] The technical solution of the embodiment of the present invention obtains the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected profit of the recommended users for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected profit is predicted based on the target product expected profit prediction model; based on the current product recommendation strategy matrix, the current conversion probability and the current expected profit, the recommendation strategy score function corresponding to each product to be recommended is determined, and by analyzing the current recommendation demand of each product to be recommended, the constraint conditions corresponding to the recommendation strategy score function are determined; based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, the current recommendation strategy score corresponding to the current product recommendation strategy matrix is ​​determined, and based on the current recommendation strategy score and the preset convergence conditions, the target product recommendation strategy is uniformly determined for all products to be recommended, so as to accurately and conveniently determine the target product recommendation strategy corresponding to all products to be recommended in the current recommendation cycle, and improve the efficiency and accuracy of determining the target product recommendation strategy, so that the products to be recommended can be effectively recommended to the users to be recommended according to the target product recommendation strategy, so as to improve user perception and company benefits.

[0092] Based on the above technical solution, the information acquisition module 310 may include:

[0093] The current product recommendation strategy matrix construction submodule is used to obtain the number of products to be recommended and the number of users to be recommended in the current recommendation cycle, and based on the number of products and the number of users, construct the current product recommendation strategy matrix of the products to be recommended in the current recommendation cycle; wherein the elements in the current product recommendation strategy matrix are used to represent the current recommendation status of the users to be recommended and the products to be recommended.

[0094] On the basis of the above technical scheme, the information acquisition module 310 is specifically used to: obtain the current user information of the user to be recommended and the current product information of the product to be recommended in the current recommendation cycle; input the current user information and the current product information into the target product conversion probability prediction model for probability prediction, and determine the current conversion probability of the user to be recommended for the product to be recommended in the current recommendation cycle based on the output of the target product conversion probability prediction model; the target product conversion probability prediction model is pre-trained based on a preset classification model; input the current user information and the current product information into the target product expected profit prediction model for profit prediction, and determine the current expected profit of the user to be recommended for the product to be recommended in the current recommendation cycle based on the output of the target product expected profit prediction model; the target product expected profit prediction model is pre-trained based on a preset regression model.

[0095] Based on the above technical solution, the recommendation strategy score function determination module 320 is specifically used to: multiply the current conversion probability by the current expected benefit to obtain the expected benefit of the recommendation of the recommended user for the recommended product; multiply each element in the current product recommendation strategy matrix by the expected benefit of the recommendation corresponding to each element to obtain the recommendation score numerator corresponding to each element; add the recommendation score numerators corresponding to each element to obtain the recommendation strategy score functions corresponding to each product to be recommended.

[0096] Based on the above technical solution, the constraints include at least one of the following: the sum of the element values ​​of each row of the current product recommendation strategy matrix is ​​one, and each element is zero or one; the sum of the element values ​​of each column of the current product recommendation strategy matrix is ​​within the preset recommendation quantity range in the current recommendation demand; the elements in the current product recommendation strategy matrix whose current conversion probability is less than or equal to the preset conversion probability in the current recommendation demand are set to zero; the elements in the current product recommendation strategy matrix whose current expected return is less than or equal to the preset expected return in the current recommendation demand are set to zero.

[0097] Based on the above technical solution, the target product recommendation strategy determination module 330 may include:

[0098] The current product recommendation strategy matrix iteration submodule is used to iteratively adjust the current product recommendation strategy matrix based on preset adjustment rules and constraints to obtain the current product recommendation strategy matrix after element values ​​are adjusted;

[0099] A current recommendation strategy score determination submodule is used to determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix after the element value is adjusted based on the current product recommendation strategy matrix after the element value is adjusted and the recommendation strategy score function;

[0100] The target product recommendation strategy determination submodule is used to determine the target product recommendation strategy corresponding to each to-be-recommended product based on the element value of each element in the current product recommendation strategy matrix when the current recommendation strategy score meets the preset convergence condition.

[0101] On the basis of the above technical scheme, the current product recommendation strategy matrix iteration submodule is specifically used for: dividing the current product recommendation strategy matrix based on the preset number of user subgroups to obtain the preset number of user subgroups to be selected; determining the elements to be adjusted in the target user subgroup based on the preset crossover probability, the preset mutation probability and the preset selection probability corresponding to each element in each user subgroup to be selected; the preset selection probability is determined based on the recommendation score numerator function corresponding to each element in each user subgroup to be selected; setting the element value of the element to be adjusted to zero or one to obtain the current product recommendation strategy matrix after the element value is adjusted.

[0102] The product recommendation strategy determination device provided in the embodiment of the present invention can execute the product recommendation strategy determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the product recommendation strategy determination method.

[0103] It is worth noting that in the embodiment determined by the above-mentioned product recommendation strategy, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0104] Embodiment 4

[0105] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0106] like Figure 4As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0107] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0108] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a product recommendation strategy determination method.

[0109] In some embodiments, the product recommendation strategy determination method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the product recommendation strategy determination method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the product recommendation strategy determination method in any other appropriate manner (e.g., by means of firmware).

[0110] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0111] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0112] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0113] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0114] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0115] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0116] An embodiment of the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements a product recommendation strategy determination method as provided in any embodiment of the present application.

[0117] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, and the programming language includes an object-oriented programming language, such as Java, Smalltalk, C++, and also includes a conventional procedural programming language, such as "C" language or similar programming language. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). The program product belongs to the same inventive concept as the product recommendation strategy determination method disclosed in each embodiment of the present application, and therefore is not repeated here.

[0118] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0119] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for determining a product recommendation strategy, characterized in that: include: Obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected return of the recommended user for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected return is predicted based on the target product expected return prediction model; Based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, determine the recommendation strategy score function corresponding to each to-be-recommended product, and determine the constraint condition corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each to-be-recommended product; Based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, the current recommendation strategy score corresponding to the current product recommendation strategy matrix is ​​determined, and based on the current recommendation strategy score and the preset convergence condition, the target product recommendation strategy corresponding to each product to be recommended is determined.

2. The method according to claim 1, characterized in that: The step of obtaining the current product recommendation strategy matrix of the products to be recommended in the current recommendation cycle includes: Obtaining the number of products to be recommended and the number of users to be recommended in the current recommendation cycle, and constructing a current product recommendation strategy matrix for the products to be recommended in the current recommendation cycle based on the number of products and the number of users; The elements in the current product recommendation strategy matrix are used to represent the current recommendation status of the to-be-recommended user and the to-be-recommended product.

3. The method according to claim 1, characterized in that Get the current conversion probability and current expected revenue of each recommended product for the recommended user in the current recommendation cycle, including: Obtain the current user information of the user to be recommended and the current product information of the product to be recommended in the current recommendation cycle; Inputting the current user information and the current product information into a target product conversion probability prediction model for probability prediction, and determining the current conversion probability of the to-be-recommended user for the to-be-recommended product within the current recommendation cycle based on the output of the target product conversion probability prediction model; the target product conversion probability prediction model is pre-trained based on a preset classification model; The current user information and the current product information are input into a target product expected revenue prediction model for revenue prediction, and based on the output of the target product expected revenue prediction model, the current expected revenue of the recommended user for the product to be recommended within the current recommendation cycle is determined; the target product expected revenue prediction model is pre-trained based on a preset regression model.

4. The method according to claim 1, characterized in that: The determining of the recommendation strategy score function corresponding to each to-be-recommended product based on the current product recommendation strategy matrix, the current conversion probability and the current expected return includes: Multiplying the current conversion probability by the current expected benefit to obtain the expected benefit of the recommended user for the recommended product; Multiply each element in the current product recommendation strategy matrix by the expected benefit of recommendation corresponding to each element to obtain a numerator function of recommendation corresponding to each element; The recommendation strategy score function corresponding to each product to be recommended is obtained by adding the recommendation score subfunctions corresponding to each element.

5. The method according to claim 1, characterized in that The constraint conditions include at least one of the following: the sum of the element values ​​of each row of the current product recommendation strategy matrix is ​​one, and each element is zero or one; the sum of the element values ​​of each column of the current product recommendation strategy matrix is ​​within the preset recommendation quantity range in the current recommendation demand; the elements in the current product recommendation strategy matrix whose current conversion probability is less than or equal to the preset conversion probability in the current recommendation demand are set to zero; the elements in the current product recommendation strategy matrix whose current expected return is less than or equal to the preset expected return in the current recommendation demand are set to zero.

6. The method according to claim 1, characterized in that The step of determining a current recommendation strategy score corresponding to the current product recommendation strategy matrix based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint condition, and determining a target product recommendation strategy corresponding to each to-be-recommended product based on the current recommendation strategy score and a preset convergence condition, includes: Iteratively adjust the current product recommendation strategy matrix based on the preset adjustment rules and the constraint conditions to obtain the current product recommendation strategy matrix after element values ​​are adjusted; Based on the current product recommendation strategy matrix after the element value is adjusted and the recommendation strategy score function, determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix after the element value is adjusted; When the current recommendation strategy score meets the preset convergence condition, the target product recommendation strategy corresponding to each to-be-recommended product is determined based on the element value of each element in the current product recommendation strategy matrix.

7. The method according to claim 6, characterized in that The iterative adjustment of the current product recommendation strategy matrix based on the preset adjustment rules and the constraint conditions to obtain the current product recommendation strategy matrix after element values ​​are adjusted includes: Dividing the current product recommendation strategy matrix based on a preset number of user subgroups to obtain a preset number of user subgroups to be selected; Determine the element to be adjusted in the target user subgroup based on the preset crossover probability, the preset mutation probability and the preset selection probability corresponding to each element in each candidate user subgroup; the preset selection probability is determined based on the recommended molecular function corresponding to each element in each candidate user subgroup; The element value of the element to be adjusted is set to zero or one to obtain the current product recommendation strategy matrix after the element value is adjusted.

8. A device for determining a product recommendation strategy, characterized in that: The device comprises: An information acquisition module is used to obtain the current product recommendation strategy matrix and current recommendation demand of the products to be recommended in the current recommendation cycle, as well as the current conversion probability and current expected return of the recommended user for each product to be recommended; the current conversion probability is predicted based on the target product conversion probability prediction model; the current expected return is predicted based on the target product expected return prediction model; A recommendation strategy score function determination module is used to determine the recommendation strategy score function corresponding to each to-be-recommended product based on the current product recommendation strategy matrix, the current conversion probability and the current expected return, and determine the constraint conditions corresponding to the recommendation strategy score function by analyzing the current recommendation demand of each to-be-recommended product; The target product recommendation strategy determination module is used to determine the current recommendation strategy score corresponding to the current product recommendation strategy matrix based on the current product recommendation strategy matrix, the recommendation strategy score function and the constraint conditions, and to determine the target product recommendation strategy corresponding to each product to be recommended based on the current recommendation strategy score and the preset convergence condition.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the product recommendation strategy determination method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the product recommendation strategy determination method as described in any one of claims 1-7 is implemented.

Citation Information

Patent Citations

  • Product recommendation method and device based on deep learning, equipment and storage medium

    CN113076484A

  • Product recommendation method and device, electronic equipment and readable storage medium

    CN115271866A

  • Strategy recommendation method and device, equipment and storage medium

    CN115357809A

  • Product recommendation method and device, equipment and storage medium

    CN116128607A

  • Service recommendation method and device, electronic equipment and storage medium

    CN117493626A