User operation strategy recommendation method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a user operation strategy recommendation method and device, computer equipment and a storage medium. The user operation strategy recommendation method comprises the following steps: classifying all to-be-recommended users according to an obtained target operation index and user attribute data to obtain a plurality of to-be-recommended user groups; performing data processing based on the user attribute data of all the to-be-recommended users in the to-be-recommended user group through a preset strategy recommendation model corresponding to each to-be-recommended user group to obtain an initial operation strategy corresponding to the to-be-recommended user group; determining an objective function, a decision variable and a constraint condition of a strategy optimization model according to the initial operation strategy, and determining an optimized operation strategy of each user group to be recommended according to a result obtained by solving the strategy optimization model; and determining target operation strategy parameters of all the to-be-recommended users according to the optimized operation strategy. According to the invention, the rationality of the operation strategy is improved, and the overall operation effect is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a method, device, computer device and storage medium for recommending user operation strategies. Background Art

[0002] User operation is a bridge connecting enterprises and users. Through user operation, enterprises can better understand user needs and thus promote business development. The behaviors and needs of users show a high degree of diversity and complexity. With the development of big data and artificial intelligence technologies, more and more enterprises hope to improve user conversion rates and satisfaction through more scientific methods. However, current user operation methods still have many deficiencies in dealing with complex user groups and optimizing the allocation of marketing resources.

[0003] In the prior art, traditional user operation methods often rely on experience and simple statistical analysis, resulting in rough user group division results, inaccurate operation strategies for different user groups, and inability to meet the personalized needs of different user groups, leading to poor conversion effects of users under different marketing means. At the same time, in the case of limited marketing resources for user operation, traditional methods are difficult to achieve the optimal allocation of marketing resources, resulting in unreasonable distribution of operation strategies, further leading to resource waste and poor operation effects. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer device and storage medium for recommending user operation strategies to solve the problems of inaccurate strategy recommendation and unreasonable distribution of operation strategies in existing user operation methods.

[0005] A method for recommending user operation strategies includes: Obtain target operation indicators and user attribute data of all users to be recommended, and classify all the users to be recommended according to the target operation indicators and the user attribute data to obtain multiple user groups to be recommended; Through a preset strategy recommendation model corresponding to each user group to be recommended, perform data processing based on the user attribute data of all users to be recommended in the user group to be recommended, and obtain an initial operation strategy corresponding to the user group to be recommended; Determine the objective function, decision variables and constraint conditions of the strategy optimization model according to the initial operation strategy; Perform a solution process on the strategy optimization model based on the objective function, decision variables and constraint conditions, and determine the optimized operation strategies of each user group to be recommended according to the solution process results; Determine the target operation strategy parameters of all the users to be recommended according to the optimized operation strategies.

[0006] A user operation strategy recommendation device, comprising: A user classification module, configured to obtain target operation metrics and user attribute data of all users to be recommended, classify all the users to be recommended according to the target operation metrics and the user attribute data, and obtain multiple groups of users to be recommended; A strategy recommendation module, configured to perform data processing on the user attribute data of all users to be recommended within each group of users to be recommended through a preset strategy recommendation model corresponding to each group of users to be recommended, and obtain an initial operation strategy corresponding to the group of users to be recommended; An optimization model establishment module, configured to determine an objective function, decision variables, and constraint conditions of a strategy optimization model according to the initial operation strategy; An optimization model solving module, configured to perform a solving process on the strategy optimization model based on the objective function, decision variables, and constraint conditions, and determine an optimized operation strategy for each group of users to be recommended according to the solving result; A target strategy determination module, configured to determine target operation strategy parameters for all the users to be recommended according to the optimized operation strategy.

[0007] A computer device, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, where when the processor executes the computer-readable instructions, the above-mentioned user operation strategy recommendation method is implemented.

[0008] A computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the user operation strategy recommendation method as described above.

[0009] In the above user operation strategy recommendation method, device, computer device, and storage medium, the user operation strategy recommendation method obtains target operation metrics and user attribute data of all users to be recommended, classifies all users to be recommended based on the target operation metrics and user attribute data to obtain multiple user groups to be recommended; through a preset strategy recommendation model corresponding to each user group to be recommended, data processing is performed based on the user attribute data of all users to be recommended within the user group to be recommended to obtain an initial operation strategy corresponding to the user group to be recommended; determines the objective function, decision variables, and constraint conditions of the strategy optimization model according to the initial operation strategy; performs a solution process on the strategy optimization model based on the objective function, decision variables, and constraint conditions, and determines the optimized operation strategies for each user group to be recommended according to the solution process results; determines the target operation strategy parameters of all users to be recommended according to the optimized operation strategies. On the basis of achieving accurate user group division, the present invention can predict the conversion effects of users to be recommended under different operation strategy parameters based on preset strategy recommendation models established in advance for different user groups by machine learning, improving the accuracy of the initial operation strategy recommendation. At the same time, the present invention uses the model recommendation results at the user group dimension as the initial operation strategy, determines the objective function, decision variables, and constraint conditions of the strategy optimization model based on the initial operation strategy, and uses a mathematical programming model to achieve the optimal allocation of marketing resources on the basis of considering the conversion effects of all users to be recommended, improving the rationality of the operation strategy allocation and improving the overall operation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts.

[0011] Figure 1 is a flowchart of a user operation strategy recommendation method in an embodiment of the present invention; Figure 2 is a flowchart of step S10 in the user operation strategy recommendation method in an embodiment of the present invention; Figure 3 is another flowchart of step S10 in the user operation strategy recommendation method in an embodiment of the present invention; Figure 4 is another flowchart of step S10 in the user operation strategy recommendation method in an embodiment of the present invention; Figure 5 is a flowchart of step S20 in the user operation strategy recommendation method in an embodiment of the present invention; Figure 6 It is another process schematic diagram of step S20 in the user operation strategy recommendation method according to an embodiment of the present invention; Figure 7 It is a process schematic diagram of step S30 in the user operation strategy recommendation device according to an embodiment of the present invention; Figure 8 It is a structural schematic diagram of the user operation strategy recommendation device according to an embodiment of the present invention; Figure 9 It is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed implementation manners

[0012] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0013] The user operation strategy recommendation method provided in this embodiment can be applied to the application scenario where product merchants or service providers launch operation strategies for users in need of services. The operation strategies can be commodity recommendation strategies, advertisement placement strategies, preferential activity planning, etc. For example, the service provider pushes different preferential activities to different registered users, so that more registered users finally purchase the service. At this time, whether to purchase the service can be used as a standard to measure the effect of the operation strategy. In actual applications, when a new service needs to be launched or the price of an already launched service is adjusted, the operation service end will trigger the process of the user operation strategy recommendation method. The operation service end can be implemented by an independent server or a server cluster composed of multiple servers.

[0014] In one embodiment, as Figure 1 shown, a user operation strategy recommendation method is provided, including the following steps S10 - S50: S10. Obtain the target operation metrics and the user attribute data of all users to be recommended, and classify all the users to be recommended according to the target operation metrics and the user attribute data to obtain multiple user groups to be recommended.

[0015] Understandably, when triggering the process of the user operation strategy recommendation method, it is necessary to first determine the users to be recommended and the target operation indicators. The users to be recommended refer to all users who need to be covered by the operation strategy, that is, the population that potentially receives the recommended content or services. The number and scope of the users to be recommended can be adjusted according to different service requirements and business goals. For example, in an e-commerce platform, the users to be recommended can include all registered users or only users in a specific region. Operation indicators refer to a series of quantitative criteria set by an enterprise during operation to measure and evaluate the achievement of specific operation activities or business goals. The target operation indicator refers to the criterion selected from multiple operation indicators to measure the effect of the current operation strategy. For example, when an enterprise promotes services to users, the target operation indicator can be the user business conversion rate reflected by users purchasing services, or the user activity rate reflected by the proportion of users logging in to their accounts. The target operation indicator is closely related to the enterprise's strategic goals, business requirements, and market competition environment, and can help the enterprise better understand the operation status and guide the formulation and optimization of decisions.

[0016] The user attribute data of the users to be recommended is a dataset used to describe the various characteristics and conditions that distinguish the users to be recommended from other users. Each user has a corresponding series of community attribute data, including data such as age, education level, occupation type, and customer acquisition channels. After the target operation indicator is determined, these user attribute data may affect the formulation of the operation strategy and the result of the target operation indicator from different aspects. Although the user attribute data between different users to be recommended is different, there is a certain degree of similarity in the user attribute data, and the degree of influence on the target operation indicator also has mutual reference value. Therefore, classifying all users to be recommended according to the target operation indicator and user attribute data can obtain multiple groups of users to be recommended. For example, when considering the influence of age on users' purchase of services, the willingness of users in different age groups to purchase services is different. Users under 40 years old are more likely to purchase services than users over 40 years old. The groups of users to be recommended are several user sets obtained by grouping the users to be recommended according to the degree of influence of different attributes on the target operation indicator. The purpose of classification processing is to divide the users to be recommended into multiple groups or clusters, so that the data points within the same group of users to be recommended are similar to each other, while the data points between different groups of users to be recommended are quite different.

[0017] S20. Through the preset strategy recommendation model corresponding to each group of users to be recommended, data processing is performed based on the user attribute data of all users to be recommended within the group of users to be recommended, and an initial operation strategy corresponding to the group of users to be recommended is obtained.

[0018] Understandably, after dividing all the users to be recommended into multiple groups of users to be recommended, by using the preset policy recommendation model corresponding to each group of users to be recommended to process the user attribute data of all the users to be recommended within the group of users to be recommended, the operation strategies of each user to be recommended can be predicted, and then the set of operation strategies of all the users to be recommended within the group of users to be recommended can be used as the initial operation strategy. The preset policy recommendation model refers to a neural network model that has been pre-trained to predict the operation strategy of a user at a specific future time (such as one week) by using the user attribute data of the user to be recommended at the current time. Since the influence degrees of the user attribute data between different groups of users to be recommended on the target operation index are different, each group of users to be recommended corresponds to a different preset policy recommendation model. The initial operation strategy refers to the set of candidate operation strategies of all users in the group of users to be recommended under different operation modes output by the preset policy recommendation model.

[0019] S30. Determine the objective function, decision variables, and constraint conditions of the policy optimization model according to the initial operation strategy.

[0020] Understandably, since the preset policy recommendation model depends on training samples, and the training samples may not conform to the operation expected target (target operation index), therefore, subsequent policy optimization needs to be carried out on the basis of the initial operation strategy. The policy optimization model is a pre-constructed optimization model for matching operation strategies for users to be recommended. In the actual business scenario, the optimal allocation problem can be solved by solving the optimization model. The policy optimization model can be a model of linear programming, non-linear programming, dynamic programming, integer programming, or system science method, preferably a linear programming model. The policy optimization model includes an objective function, decision variables, and constraint conditions. Among them, the objective function refers to the function to be extremized that has a direct relationship with the decision variables and is used to guide the solution process of the decision variables, such as the user conversion rate function. The decision variables refer to the variables to be solved in the allocation business scenario that are related to the constraint conditions and optimization objectives, such as the operation modes of the users to be recommended. The constraint conditions refer to the restrictions that the decision variables must satisfy when solving the objective function, such as the marketing budget restriction.

[0021] S40. Perform a solution process on the policy optimization model based on the objective function, decision variables, and constraint conditions, and determine the optimized operation strategies of each group of users to be recommended according to the solution result.

[0022] Understandably, when the policy optimization model is a linear programming model, an existing linear programming solver can be called to solve the policy optimization model to obtain the solution results of the decision variables that satisfy the constraint conditions and the objective function. The solver includes appropriate solution algorithms, such as the simplex method, the interior point method, etc. According to the solution results, the optimized operation strategies for each user group to be recommended can be determined. The solution results are expressed as the optimal matching results of all users to be recommended within the user group to be recommended for different operation methods, that is, the optimal marketing resource allocation plan. The optimized operation strategy refers to the distribution of different operation methods among all users in the user group to be recommended. For example, 30% of the users to be recommended in the same user group to be recommended use the telephone method to promote the business to be marketed, 40% of the users to be recommended use the method of issuing coupons to promote the business to be marketed, and 30% of the users to be recommended use the method of time-limited benefits to promote the business to be marketed.

[0023] S50. Determine the target operation strategy parameters for all the users to be recommended according to the optimized operation strategy.

[0024] Understandably, based on the optimized operation strategy, the target operation strategy parameters for all users to be recommended can be determined. The target operation strategy parameters refer to the specific operation strategy methods finally adopted by the users to be recommended and the corresponding strategy parameters. For example, the strategy name, strategy parameters, and budget allocation invested by the users to be recommended. After obtaining the target operation strategy parameters of the users to be recommended, the target operation strategy parameters are sent to the user device preset for executing the operation strategy plan. For example, which specific user to be recommended in the same user group to be recommended uses the telephone method, which uses the method of issuing coupons, and which uses the method of time-limited benefits. On the basis of considering the user conversion effects of all users to be recommended, the optimal allocation of operation resources is achieved based on the method of solving the mathematical programming model, so that the operation decision experiment period can be greatly shortened.

[0025] In this embodiment, by obtaining the target operation metrics and the user attribute data of all users to be recommended, classifying all users to be recommended according to the target operation metrics and the user attribute data, and obtaining multiple user groups to be recommended; through the preset policy recommendation model corresponding to each user group to be recommended, based on the user attribute data of all users to be recommended in the user group to be recommended, data processing is performed to obtain the initial operation policy corresponding to the user group to be recommended; according to the initial operation policy, the objective function, decision variables, and constraint conditions of the policy optimization model are determined; based on the objective function, decision variables, and constraint conditions, the policy optimization model is solved, and according to the solution result, the optimized operation policies of each user group to be recommended are determined; according to the optimized operation policies, the target operation policy parameters of all users to be recommended are determined. On the basis of realizing accurate user group division, this embodiment can predict the conversion effect of users to be recommended under different operation policy parameters based on the preset policy recommendation models established in advance for different user groups by machine learning, improving the accuracy of the initial operation policy recommendation. At the same time, this embodiment uses the model recommendation result at the user group dimension as the initial operation policy, and determines the objective function, decision variables, and constraint conditions of the policy optimization model based on the initial operation policy. By using the mathematical programming model, on the basis of considering the conversion effects of all users to be recommended, the optimal allocation of marketing resources is realized, improving the rationality of the operation policy allocation and improving the overall operation effect.

[0026] In one embodiment, as Figure 2 shown, in step S10, that is, the step of classifying all the users to be recommended according to the target operation metrics and the user attribute data to obtain multiple user groups to be recommended includes: S101. Obtain a target decision tree classification model corresponding to the target operation metrics; S102. Perform data processing on the user attribute data through the target decision tree classification model to obtain a decision tree classification result; S103. Classify all the users to be recommended into a preset number of user groups to be recommended according to the decision tree classification result.

[0027] Understandably, the target operation metrics are business goals that one hopes to optimize or achieve, such as increasing user activity rate, increasing user retention rate, and improving user conversion rate, etc. User attribute data includes various characteristic information of users, such as age, gender, and region, etc. The impact degree of the same user attribute data on different operation metrics is different. Therefore, different target operation metrics correspond to different target decision tree classification models. The target decision tree classification model is pre-trained based on historical data and business logic, and is a classification model used to establish an association relationship between user attribute data and target operation metrics. The decision tree structure in the target decision tree classification model can reflect the impact degree of different user attributes on the target operation metrics. Inputting the user attribute data into the target decision tree classification model, the model will process and analyze the data according to preset rules and logic, and obtain the decision tree classification result. The decision tree classification result is the result output by the target decision tree classification model for classifying the users to be recommended into which category according to the impact of different user attributes on the target operation metrics. Classifying all the users to be recommended into a preset number of user groups to be recommended according to the decision tree classification result, where the preset number is a fixed value corresponding to the target decision tree classification model and used to represent the number of groups after classification, for example, the default value of the preset number is set to 8.

[0028] The target decision tree classification model of this embodiment can accurately analyze the impact degree of different user attribute data on the target operation metrics, so as to divide the users to be recommended into different user groups, ensuring the rationality of user grouping and helping to improve the business effect under the guidance of the target operation metrics.

[0029] In one embodiment, as Figure 3 shown, before step S101, that is, before obtaining the target decision tree classification model corresponding to the target operation metric, it includes: S1011. Obtain the historical operation results of all the users to be recommended corresponding to the target operation metric, and the user historical attribute data corresponding to the historical operation results; S1012. Determine the target attribute features and classification sample data sets corresponding to the target operation metric according to the user historical attribute data; S1013. Perform model training on the preset decision tree model based on the target attribute features and the classification sample data sets until the number of user groups in the decision tree training result reaches the preset number, and obtain the target decision tree classification model corresponding to the target operation metric.

[0030] Understandably, before obtaining the target decision tree classification model corresponding to the target operation metric, it is necessary to train the target decision tree classification model based on sample data. Obtain the historical operation results of all users to be recommended corresponding to the target operation metric, as well as the user historical attribute data corresponding to the historical operation results, in order to determine the sample data. The historical operation result refers to the result reflecting the target operation metric among the users to be recommended for whom the service has been launched within a specific historical time period (such as 1 month). For example, when the target operation metric is user purchase of the service, the historical operation results include two operation results: the users for whom the service has been launched have purchased the service and have not purchased the service. The user historical attribute data is the attribute data that affects the historical operation result, including the attribute data of the users who have purchased the service and the attribute data of the users who have not purchased the service.

[0031] The methods for building the decision tree in the target decision tree classification model include, but are not limited to, generating the decision tree using the CART method and the LightGBM algorithm framework. Preferably, the Light GBM algorithm framework can be used to generate a single decision tree, realizing the automatic binning function for categorical features, ensuring the logical clarity of dividing the users to be recommended to the greatest extent, and having business interpretability. Determine the target attribute features and the classification sample data set corresponding to the target operation metric from the historical operation results and the user historical attribute data. The target attribute features are usually selected from the data set through specific quantitative evaluation methods. The target attribute features refer to one or more features selected from the user attributes that have a significant impact on the target operation metric, such as age features, educational background features, and occupation type features, etc. The classification sample data set refers to the sample data set used to train the decision tree.

[0032] Perform model training on the preset decision tree model based on the target attribute features and the classification sample data set until the number of user groups in the decision tree training result reaches the preset number, and obtain the target decision tree classification model corresponding to the target operation metric. During the training process, the decision tree model is a hierarchical structure, and the model will gradually build the decision tree according to the user historical attribute data until the predetermined stopping condition is reached. The stopping condition is that the number of user groups reaches the preset number, and this preset number can be set according to business requirements to ensure that the complexity and generalization ability of the decision tree are appropriate.

[0033] In one embodiment, the target operation metric is the user conversion rate reflected by users purchasing services. The target attribute features include age features, educational attainment features, and occupation type features, and the preset quantity is 8. At this time, the influence degree of age features on users purchasing services is greater than that of educational attainment features, and the influence degree of educational attainment features on users purchasing services is greater than that of occupation type features. First, the age features are determined as the classification dimension of the first level, and all users to be recommended are classified according to the age features, obtaining 2 user groups to be recommended at the first level. Secondly, since the number of user groups has not reached the preset quantity, the educational attainment features are determined as the classification dimension of the second level, and all users to be recommended are classified based on the 2 user groups to be recommended at the first level according to the educational attainment features, obtaining 4 user groups to be recommended at the second level. Finally, since the number of user groups still has not reached the preset quantity, the occupation type features are determined as the classification dimension of the third level, and all users to be recommended are classified based on the 4 user groups to be recommended at the second level according to the occupation type features, obtaining 8 user groups to be recommended at the third level. At this time, the number of user groups reaches the preset quantity, and the training of the target decision tree classification model is completed.

[0034] In another embodiment, by performing model training according to different classification methods (such as direct classification method and classification by business line method), different target decision tree classification models can be obtained. Furthermore, according to the target decision tree classification model, data processing is performed on user attribute data to obtain a decision tree classification result. When the direct classification method is adopted, the preset decision tree model is directly trained based on the target attribute features and the classification sample data set to obtain a target decision tree classification model corresponding to the target operation metric. When the classification by business line method is adopted, first, all users to be recommended are divided according to the business line (for example, in an e-commerce enterprise, there may be a clothing business line, an electronic product business line, and a household goods business line), and then the preset decision tree model is trained based on the target attribute features and the classification sample data set within each business line to obtain a target decision tree classification model corresponding to the target operation metric. In practical applications, the direct classification method is preferably adopted for model training and the trained target decision tree classification model is used for user classification.

[0035] This embodiment can construct a target decision tree classification model for a specific target operation metric, providing strong support for business decisions. At the same time, it helps to improve the classification accuracy of the trained target decision tree classification model during application.

[0036] In one embodiment, as Figure 4 shown, the user historical attribute data includes user basic attribute data and user business attribute data; in step S1012, that is, determining the target attribute features and the classification sample data set corresponding to the target operation metric according to the user historical attribute data includes: S10121. Perform user demand analysis and processing on all the user basic attribute data and user business attribute data based on the target operation index, and determine the target attribute features corresponding to the target operation index according to the results of the user demand analysis; S10122. Determine each historical operation result corresponding to the target operation index and the user basic attribute data and user business attribute data corresponding to the historical operation result as a classification sample data group corresponding to the target operation index, and determine a classification sample data set corresponding to the target operation index according to all the classification sample data groups.

[0037] Understandably, user historical attribute data includes user basic attribute data and user business attribute data. User basic attribute data refers to the features corresponding to the user's own dimension obtained with the user's authorization and consent, such as basic attribute data such as age, education level, and occupation type. User business attribute data refers to the features corresponding to the user's business behavior dimension, such as the user's registered account date, customer acquisition channel, customer group classification label, specific business identifier, and other business attribute data. Performing user demand analysis and processing on all user basic attribute data and user business attribute data based on the target operation index can determine the target attribute features corresponding to the target operation index. User demand analysis and processing refers to the process of evaluating the degree of influence of different attributes on the target operation index, that is, the process of using the actual business conversion results such as user usage rights, entering a specified business operation link, and purchasing services as sample labels to select target attribute features. The features selected when training a decision tree model are determined based on their contribution to the model performance. Through quantitative evaluation methods and feature selection steps, the most important features can be selected to build a concise and efficient decision tree model. Specifically, the CART algorithm recursively divides the data set into two subsets until the stopping condition is met. That is, using the binary recursive partitioning technique, each time the current sample set is divided into two sub-sample sets, and the generated decision tree is a binary tree with a simple structure. When constructing a decision tree using the CART algorithm, the Gini index is used as the feature selection criterion. The smaller the Gini index, the smaller the uncertainty of the data set, that is, the higher the purity of the data set. Therefore, select the feature that can minimize the Gini coefficient as the splitting criterion, and select one or more target attribute features based on the user basic attribute data and user business attribute data.

[0038] Each historical operation result corresponding to the target operation index, along with the user basic attribute data and user business attribute data corresponding to the historical operation result, is determined as a classification sample data group corresponding to the target operation index. And a classification sample data set corresponding to the target operation index is determined based on all the classification sample data groups. A classification sample data group refers to a historical operation result and all the historical attribute data that affect the historical operation result. A classification sample data set refers to the set of all classification sample data groups.

[0039] This embodiment considers the influence of multi-dimensional attributes on classification based on user basic attribute data and user business attribute data. At the same time, through feature selection, the attributes with more important influence are determined as target attribute features, which can reduce the number of features, thereby reducing the complexity of the model and improving the performance of the model, and helping to construct a concise and efficient decision tree model.

[0040] In one embodiment, as Figure 5 shown, in step S20, that is, before data processing is performed on the user attribute data of all the to-be-recommended users in the to-be-recommended user group through the preset policy recommendation model corresponding to each to-be-recommended user group, it includes: S201. Obtain the historical operation data of all the to-be-recommended users in each to-be-recommended user group, where the historical operation data includes historical operation policy parameters, user historical attribute data, and historical operation results; S202. Determine the group policy sample data of each to-be-recommended user group according to the historical operation policy parameters and user historical attribute data, and determine the sample true value corresponding to the group policy sample data according to the historical operation results; S203. Input the group policy sample data into the constructed initial policy recommendation model to obtain the sample predicted value output by the initial policy recommendation model; S204. Determine the sample error according to the sample predicted value and the sample true value, and stop training when the sample error reaches the preset precision threshold to obtain the preset policy recommendation model corresponding to each to-be-recommended user group.

[0041] Understandably, before processing data through the preset policy recommendation models corresponding to each user group to be recommended, it is necessary to first train the preset policy recommendation models corresponding to each user group to be recommended using machine learning methods. Obtain the historical operation data of all users to be recommended within each user group to be recommended. Historical operation data refers to the operation-related data of users to be recommended for whom the business has been launched within a specific historical time period (such as 1 month). Historical operation data includes not only user historical attribute data and historical operation results, but also historical operation policy parameters. Historical operation policy parameters refer to the operation methods and specific parameters adopted for specific users to be recommended in the launched business. For example, the method of issuing coupons is adopted, and the amount parameter of the coupon is 100 minus 20 when full. Historical operation results can be obtained by collecting and statistically analyzing data within a statistical cycle. For example, the sequence data and statistical data of user behaviors such as registering accounts, logging in, using rights and interests, entering specified business operation links, and purchasing services within a statistical cycle before and after the execution of a marketing action for a business can reflect operation indicators such as the number of new users, user activity rate, user retention rate, and business conversion rate. Based on the historical operation policy parameters and user historical attribute data, determine the group policy sample data for each user group to be recommended, and determine the sample true value corresponding to the group policy sample data according to the historical operation results. Group policy sample data refers to a set of sample data sets extracted for users in a specific user group to be recommended in the operation policy for testing or prediction. Group policy sample data can be used to simulate and evaluate the effects of different operation methods in the user group, so as to provide data support for subsequent model training.

[0042] Input the group policy sample data into the constructed initial policy recommendation model to obtain the sample prediction values output by the initial policy recommendation model. Each group policy sample data corresponds to a sample true value and a sample prediction value. The sample true value refers to the historical operation result corresponding to the historical operation policy parameters and user historical attribute data, and the sample prediction value refers to the possible operation result output by the model after inputting the historical operation policy parameters and user historical attribute data into the initial policy recommendation model. The initial policy recommendation model refers to the policy recommendation model before training, which can be a Gradient Boosting Decision Tree (GBDT) or a neural network algorithm model. Among them, the gradient boosting model can be selected as the Light Gradient Boosting Machine (Light GBM) or the eXtreme Gradient Boosting (XGBoost) according to needs. During the training process, divide the group policy sample data into a training set and a test set, and use the training set to train the pre-constructed gradient boosting model. The model will continuously adjust its internal parameters and structure to minimize the loss function, thereby improving the accuracy of future policy prediction.

[0043] The sample error between the sample predicted value and the sample true value is used to quantify the difference between the prediction output of the machine learning algorithm and the actual target value, that is, the loss function for evaluating the prediction performance of the neural network model based on sample data. During the training process, the gradient of the sample error loss of the backpropagation algorithm with respect to the model parameters is used for iterative training, and the training is stopped when the sample error reaches the preset accuracy threshold, and a preset policy recommendation model corresponding to each user group to be recommended is obtained. The preset accuracy threshold is a critical value of the sample error preset in advance for triggering the iteration stop, such as being set to 5%.

[0044] In this embodiment, model training is performed based on the historical operation data of all users to be recommended in each user group to be recommended, and a differentiated preset policy recommendation model can be customized for different user groups to be recommended, thereby improving the effectiveness and applicability of the preset policy recommendation model.

[0045] In one embodiment, as Figure 6 shown, in step S20, that is, through the preset policy recommendation model corresponding to each user group to be recommended, data processing is performed based on the user attribute data of all users to be recommended in this user group to be recommended, and an initial operation policy corresponding to this user group to be recommended is obtained, including: S205. Select a candidate operation policy parameter from multiple candidate operation policy parameters and determine it as the initial operation policy parameter. Through the preset policy recommendation model corresponding to each user group to be recommended, data processing is performed on the initial operation policy parameter and the user attribute data of each user to be recommended in this user group to be recommended, and candidate operation results of each user to be recommended under the initial operation policy parameter are obtained; S206. Determine whether there are candidate operation policy parameters that have not been selected; S207. When there are candidate operation policy parameters that have not been selected, return to the step of selecting a candidate operation policy parameter from multiple candidate operation policy parameters and determining it as the initial operation policy parameter until all the candidate operation policy parameters are traversed; S208. When there are no candidate operation policy parameters that have not been selected, perform data set processing on the candidate operation results of all users to be recommended in the same user group to be recommended under each initial operation policy parameter, and obtain the initial operation policy of each user group to be recommended.

[0046] Understandably, when processing data through a preset policy recommendation model corresponding to each user group to be recommended, it is necessary to predict the user attribute data of the users to be recommended according to different operation methods, so as to obtain the operation results of each user to be recommended under different operation methods. Candidate operation strategy parameters refer to various different operation means and corresponding operation parameters that are preset for selection in the process of formulating operation strategies in order to achieve specific operation goals, such as the method of telemarketing, the method of issuing coupons, or the method of limited-time benefits. Initial operation strategy parameters refer to the candidate operation strategy parameters selected as the input of the preset policy recommendation model in a round of prediction. In order to traverse the candidate operation strategy parameters, multiple rounds of prediction processes are required. Candidate operation results refer to the operation results predicted by the preset policy recommendation model when the users to be recommended adopt the candidate operation strategy parameters. For example, when a specific service is pushed to the users to be recommended by the method of issuing coupons, whether the users to be recommended will purchase the service.

[0047] Select a candidate operation strategy parameter from multiple candidate operation strategy parameters. Through the preset policy recommendation model corresponding to each user group to be recommended, data processing is performed on the selected candidate operation strategy parameter (initial operation strategy parameter) and the user attribute data of each user to be recommended in the user group to be recommended, so as to obtain the candidate operation results of each user to be recommended under the selected candidate operation strategy parameter (initial operation strategy parameter), and then judge whether there are still unselected candidate operation strategy parameters. Specifically, select the method of issuing coupons as the operation strategy and the coupon parameter is a reduction of 100 yuan when the total amount reaches 200 yuan. The preset policy recommendation model corresponds to the user group to be recommended whose age is under 40 years old, education level is undergraduate or above, and occupation type is unit employee. Through this preset policy recommendation model, data processing is performed on the selected candidate operation strategy parameter and the user attribute data of each user to be recommended in the user group to be recommended, and data processing is also performed using the preset policy recommendation model corresponding to other user groups to be recommended, so as to obtain the results of whether the users to be recommended in all user groups to be recommended will purchase the service under the method of issuing coupons. Next, judge whether there are still unselected candidate operation strategy parameters, such as whether the method of limited-time benefits can still be selected as the operation strategy and the limited-time benefit parameter is to give gifts, until all candidate operation strategy parameters are traversed. Finally, data set processing is performed on the candidate operation results of all users to be recommended in the same user group to be recommended under each candidate operation strategy parameter, so as to obtain the initial operation strategy of each user group to be recommended. The initial operation strategy of the user group to be recommended shows the operation results of each user to be recommended in the user group to be recommended under various operation methods.

[0048] In this embodiment, the preset policy recommendation model traverses all candidate operation policy parameters to predict the operation results, ensuring the comprehensiveness of policy recommendation, so as to subsequently find a more reasonable operation policy for each user group to be recommended by comparing the operation effects under different operation policy parameters.

[0049] In one embodiment, as Figure 7 shown, in step S30, that is, determining the objective function, decision variables and constraint conditions of the policy optimization model according to the initial operation policy, includes: S301. Obtain the optimization target information corresponding to the initial operation policy, and determine the objective function of the policy optimization model according to the optimization target information; S302. Obtain the policy ratio information of each user group to be recommended in the initial operation policy, and determine the decision variables of the policy optimization model according to the policy ratio information; S303. Obtain the operation resource parameters corresponding to the initial operation policy, and determine the constraint conditions of the policy optimization model according to the operation resource parameters.

[0050] Understandably, since the initial operation policy shows the operation results of each user to be recommended in the user group to be recommended under various operation methods, and finally a user to be recommended can only adopt one operation method, it is necessary to decide what strategy to use and what the specific settings of the strategy parameters are, that is, to determine the objective function, decision variables and constraint conditions of the policy optimization model. The optimization target information refers to the maximum or minimum conditions that the operation activities of the users to be recommended need to achieve. The objective function can correspond to only one optimization target information, or can correspond to multiple optimization target information at the same time. The process of determining the objective function of the policy optimization model according to the optimization target information is the process of converting the optimization target information into the function expression form of the decision variables in the policy optimization model, and the objective function is the function expression form corresponding to the optimization target information. For example, when the optimization target information is to maximize the number of user conversions, it means that the operation policy needs to deliver a suitable operation method to each user to be recommended, so that the number of users who purchase services within a statistical period reaches the maximum. The policy ratio information of the user group to be recommended refers to the proportion of users in the user group to be recommended who adopt the operation methods corresponding to the candidate operation policy parameters in the total number of users. Specifically, the policy ratio information is the proportion of the number of users in a user group who adopt different operation methods corresponding to the candidate operation policy parameters. The operation resource parameters refer to the parameters used to represent the available marketing resource quantity, such as the total budget of all operation methods.

[0051] In one embodiment, assume that there are user groups to be recommended, operation methods, and each operation method is applicable to the users in each user group to be recommended. represents the serial number of the user group to be recommended, represents the serial number of the operation mode. For the -th user group to be recommended, the number of users is , and for the j-th operation mode, the unit cost is , the available amount of resources is , for the -th user group, if the -th operation mode is adopted, the user conversion rate is , and the total budget for operation resources is . The conversion of a single user has a large randomness, and the model prediction error is more prominent. While the conversion of a user group has relatively less randomness, and the model prediction error is relatively less obvious. Therefore, the decision variable is , which represents the proportion of users in the -th user group who are marketed using the -th operation mode. The objective function is , indicating to maximize the number of user conversions. The constraint conditions include: the proportion of users in the -th user group who are marketed using the -th operation mode is 0% - 100%, that is ; the sum of the proportions of users in the -th user group who are marketed using different operation modes is 100%, that is ; the total number of users using the -th operation mode cannot exceed the available amount of this operation mode, that is ; the marketing cost cannot exceed the total budget, that is . Specifically, the strategy optimization model obtained based on the objective function, decision variables, and constraint conditions is as follows: s.t. .

[0052] In this embodiment, a strategy optimization model is constructed based on the objective function, decision variables, and constraint conditions, ensuring that the strategy optimization process is directed towards the optimization goal and at the same time conforms to the limiting conditions in actual operation, reflecting the pertinence of the strategy optimization. Among them, the decision variables are for the proportions of users adopting various operation modes in each user group to be recommended, rather than for a single user to be recommended, which can reduce the influence of the prediction error caused by the preset strategy recommendation model and improve the rationality of the optimized operation strategy.

[0053] In one embodiment, a comparative experiment is set up to test the effectiveness of the user operation strategy recommendation method, providing a basis for optimizing the business marketing strategy. The comparative experiment method includes, but is not limited to, randomly dividing all the users to be recommended into an experimental group and a control group according to a specified ratio. For example, after a 1:1 division, half are in the experimental group and half are in the control group. During a statistical period, the users to be recommended in the experimental group are operated using the user operation strategy recommendation method in the present invention, and the conversion rate of the users who purchase the service in the experimental group is recorded. At the same time, the users to be recommended in the control group are operated using an existing unified operation method (such as uniformly using the method of issuing coupons), and the conversion rate of the users who purchase the service in the control group is recorded. Finally, by comparing the user conversion rates of the experimental group and the control group, if the user conversion rate of the experimental group is greater than that of the control group, it indicates that the user operation strategy recommendation method in the present invention is superior to the existing user operation plan, and the user operation strategy recommendation method can be applied to subsequent user operations. This embodiment verifies the effectiveness of the user operation strategy recommendation method through scientific comparative tests, ensuring the accuracy and effectiveness of user operations.

[0054] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not imply the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0055] In one embodiment, a user operation strategy recommendation device is provided, and this user operation strategy recommendation device corresponds one-to-one with the user operation strategy recommendation method in the above embodiment. As Figure 8 shown, this user operation strategy recommendation device includes a user classification module 10, a strategy recommendation module 20, an optimization model establishment module 30, an optimization model solution module 40, and a target strategy determination module 50. The detailed description of each functional module is as follows: The user classification module 10 is used to obtain the target operation index and the user attribute data of all the users to be recommended, and classify all the users to be recommended according to the target operation index and the user attribute data to obtain multiple groups of users to be recommended; The strategy recommendation module 20 is used to perform data processing on the user attribute data of all the users to be recommended in each group of users to be recommended through a preset strategy recommendation model corresponding to each group of users to be recommended, and obtain an initial operation strategy corresponding to each group of users to be recommended; The optimization model establishment module 30 is used to determine the objective function, decision variables, and constraint conditions of the strategy optimization model according to the initial operation strategy; The optimization model solution module 40 is used to perform solution processing on the strategy optimization model based on the objective function, decision variables, and constraint conditions, and determine the optimized operation strategies of each group of users to be recommended according to the solution processing results; A target policy determination module 50, configured to determine target operation policy parameters for all the to-be-recommended users according to the optimized operation policy.

[0056] In one embodiment, the user classification module 10 includes: A classification model determination unit, configured to obtain a target decision tree classification model corresponding to the target operation index; A classification result acquisition unit, configured to perform data processing on the user attribute data through the target decision tree classification model to obtain a decision tree classification result; A to-be-recommended user group determination unit, configured to classify all the to-be-recommended users into a preset number of to-be-recommended user groups according to the decision tree classification result.

[0057] In one embodiment, the user classification module 10 further includes: A historical data acquisition unit, configured to obtain the historical operation results of all the to-be-recommended users corresponding to the target operation index, and the user historical attribute data corresponding to the historical operation results; A historical data analysis unit, configured to determine a target attribute feature and a classification sample data set corresponding to the target operation index according to the user historical attribute data; A decision tree model training unit, configured to perform model training on a preset decision tree model based on the target attribute feature and the classification sample data set until the number of user groups in the decision tree training result reaches the preset number, so as to obtain a target decision tree classification model corresponding to the target operation index.

[0058] In one embodiment, the user classification module 10 further includes: A target attribute feature determination unit, configured to perform user demand analysis and processing on all the user basic attribute data and user service attribute data based on the target operation index, and determine a target attribute feature corresponding to the target operation index according to the user demand analysis result; A classification sample data set determination unit, configured to determine each historical operation result corresponding to the target operation index, and the user basic attribute data and user service attribute data corresponding to the historical operation result as a classification sample data group corresponding to the target operation index, and determine a classification sample data set corresponding to the target operation index according to all the classification sample data groups.

[0059] In one embodiment, the policy recommendation module 20 includes: A historical operation data acquisition unit, configured to obtain the historical operation data of all the to-be-recommended users in each to-be-recommended user group, where the historical operation data includes historical operation policy parameters, user historical attribute data, and historical operation results; A sample determination unit, configured to determine the group policy sample data of each of the to-be-recommended user groups according to the historical operation policy parameters and the user historical attribute data, and determine the sample true value corresponding to the group policy sample data according to the historical operation result; A sample prediction value determination unit, configured to input the group policy sample data into the constructed initial policy recommendation model, and obtain the sample prediction value output by the initial policy recommendation model; A model training unit, configured to determine the sample error according to the sample prediction value and the sample true value, stop training when the sample error reaches a preset accuracy threshold, and obtain a preset policy recommendation model corresponding to each of the to-be-recommended user groups.

[0060] In an embodiment, the policy recommendation module 20 further includes: A user policy recommendation unit, configured to select a candidate operation policy parameter from multiple candidate operation policy parameters as the initial operation policy parameter, and perform data processing on the initial operation policy parameter and the user attribute data of each to-be-recommended user in the to-be-recommended user group through the preset policy recommendation model corresponding to each to-be-recommended user group, so as to obtain the candidate operation results of each to-be-recommended user under the initial operation policy parameter; A parameter selection judgment unit, configured to judge whether there are candidate operation policy parameters that have not been selected; A parameter selection traversal unit, configured to, when there are candidate operation policy parameters that have not been selected, return to the step of selecting a candidate operation policy parameter from multiple candidate operation policy parameters as the initial operation policy parameter until all the candidate operation policy parameters are traversed; An initial operation policy determination unit, configured to, when there are no candidate operation policy parameters that have not been selected, perform data set processing on the candidate operation results of all to-be-recommended users in the same to-be-recommended user group under each initial operation policy parameter, so as to obtain the initial operation policy of each to-be-recommended user group.

[0061] In an embodiment, the optimization model establishment module 30 includes: A target function determination unit, configured to obtain the optimization target information corresponding to the initial operation policy, and determine the target function of the policy optimization model according to the optimization target information; A decision variable determination unit, configured to obtain the policy ratio information of each to-be-recommended user group in the initial operation policy, and determine the decision variable of the policy optimization model according to the policy ratio information; A constraint condition determination unit, configured to obtain the operation resource parameters corresponding to the initial operation policy, and determine the constraint conditions of the policy optimization model according to the operation resource parameters.

[0062] For the specific limitations of the user operation strategy recommendation device, reference can be made to the limitations of the user operation strategy recommendation method in the foregoing text, which will not be elaborated here. Each module in the above user operation strategy recommendation device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0063] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 9 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The database of the computer device is used to store the data involved in the user operation strategy recommendation method. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a user operation strategy recommendation method is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0064] In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor. When the processor executes the computer-readable instructions, the following steps are implemented: Obtain the target operation metrics and the user attribute data of all users to be recommended, classify all the users to be recommended according to the target operation metrics and the user attribute data, and obtain multiple groups of users to be recommended; Through a preset policy recommendation model corresponding to each group of users to be recommended, perform data processing based on the user attribute data of all users to be recommended within the group of users to be recommended, and obtain an initial operation policy corresponding to the group of users to be recommended; Determine the objective function, decision variables, and constraint conditions of the policy optimization model according to the initial operation policy; Perform a solution process on the policy optimization model based on the objective function, decision variables, and constraint conditions, and determine the optimized operation policies for each group of users to be recommended according to the solution process results; Determine the target operation policy parameters for all users to be recommended according to the optimized operation policies.

[0065] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. Computer-readable instructions are stored on the readable storage media. When the computer-readable instructions are executed by one or more processors, the following steps are implemented: Obtain target operation metrics and user attribute data of all users to be recommended, classify all the users to be recommended according to the target operation metrics and the user attribute data, and obtain multiple groups of users to be recommended; Through a preset policy recommendation model corresponding to each group of users to be recommended, perform data processing based on the user attribute data of all users to be recommended within the group of users to be recommended, and obtain an initial operation policy corresponding to the group of users to be recommended; Determine the objective function, decision variables, and constraint conditions of the policy optimization model according to the initial operation policy; Perform a solution process on the policy optimization model based on the objective function, decision variables, and constraint conditions, and determine the optimized operation policies of each group of users to be recommended according to the solution process results; Determine the target operation policy parameters of all the users to be recommended according to the optimized operation policies.

[0066] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing related hardware through computer-readable instructions. The computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0067] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0068] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for recommending user operation strategies, characterized in that, Including: Obtain the target operation metrics and the user attribute data of all users to be recommended, classify all the users to be recommended according to the target operation metrics and the user attribute data, and obtain multiple groups of users to be recommended; Through the preset policy recommendation model corresponding to each group of users to be recommended, perform data processing based on the user attribute data of all users to be recommended within the group of users to be recommended, and obtain the initial operation policy corresponding to the group of users to be recommended; Determine the objective function, decision variables, and constraint conditions of the policy optimization model according to the initial operation policy; Perform a solution process on the policy optimization model based on the objective function, decision variables, and constraint conditions, and determine the optimized operation policies for each group of users to be recommended according to the solution result; Determine the target operation policy parameters of all the users to be recommended according to the optimized operation policy.

2. The user operation strategy recommendation method according to claim 1, wherein The classifying all the users to be recommended according to the target operation metrics and the user attribute data to obtain multiple groups of users to be recommended includes: Obtain the target decision tree classification model corresponding to the target operation metrics; Perform data processing on the user attribute data through the target decision tree classification model to obtain a decision tree classification result; Classify all the users to be recommended into a preset number of groups of users to be recommended according to the decision tree classification result.

3. The user operation strategy recommendation method according to claim 2, wherein, Before obtaining the target decision tree classification model corresponding to the target operation metrics, it includes: Obtain the historical operation results of all the users to be recommended corresponding to the target operation metrics, and the user historical attribute data corresponding to the historical operation results; Determine the target attribute features and classification sample data sets corresponding to the target operation metrics according to the user historical attribute data; Perform model training on the preset decision tree model based on the target attribute features and the classification sample data sets until the number of user groups in the decision tree training result reaches the preset number, and obtain the target decision tree classification model corresponding to the target operation metrics.

4. The user operation strategy recommendation method according to claim 3, wherein The user historical attribute data includes user basic attribute data and user business attribute data; The determining the target attribute features and classification sample data sets corresponding to the target operation metrics according to the user historical attribute data includes: Perform user demand analysis processing on all the user basic attribute data and user business attribute data based on the target operation metrics, and determine the target attribute features corresponding to the target operation metrics according to the user demand analysis result; Determine each historical operation result corresponding to the target operation metrics and the user basic attribute data and user business attribute data corresponding to the historical operation result as a classification sample data group corresponding to the target operation metrics, and determine the classification sample data set corresponding to the target operation metrics according to all the classification sample data groups.

5. The user operation strategy recommendation method according to claim 1, wherein, Before performing data processing on the user attribute data of all users to be recommended within each group of users to be recommended through the preset policy recommendation model corresponding to each group of users to be recommended, it includes: Obtain the historical operation data of all the to-be-recommended users in each of the to-be-recommended user groups, where the historical operation data includes historical operation strategy parameters, user historical attribute data, and historical operation results; Determine the group strategy sample data of each of the to-be-recommended user groups according to the historical operation strategy parameters and the user historical attribute data, and determine the sample true value corresponding to the group strategy sample data according to the historical operation results; Input the group strategy sample data into the constructed initial strategy recommendation model, and obtain the sample prediction value output by the initial strategy recommendation model; Determine the sample error according to the sample prediction value and the sample true value, and stop training when the sample error reaches the preset precision threshold to obtain the preset strategy recommendation model corresponding to each of the to-be-recommended user groups.

6. The user operation strategy recommendation method according to claim 1, wherein Through the preset strategy recommendation model corresponding to each of the to-be-recommended user groups, perform data processing based on the user attribute data of all the to-be-recommended users in the to-be-recommended user group to obtain the initial operation strategy corresponding to the to-be-recommended user group, including: Select one candidate operation strategy parameter from multiple candidate operation strategy parameters and determine it as the initial operation strategy parameter. Through the preset strategy recommendation model corresponding to each of the to-be-recommended user groups, perform data processing on the initial operation strategy parameter and the user attribute data of each of the to-be-recommended users in the to-be-recommended user group to obtain the candidate operation results of each of the to-be-recommended users under the initial operation strategy parameter; Judge whether there are unselected candidate operation strategy parameters; When there are unselected candidate operation strategy parameters, return to the step of selecting one candidate operation strategy parameter from multiple candidate operation strategy parameters and determining it as the initial operation strategy parameter until all the candidate operation strategy parameters are traversed; When there are no unselected candidate operation strategy parameters, perform data set processing on the candidate operation results of all the to-be-recommended users in the same to-be-recommended user group under each of the initial operation strategy parameters to obtain the initial operation strategy of each of the to-be-recommended user groups.

7. The user operation strategy recommendation method according to claim 1, characterized in that The determination of the objective function, decision variables, and constraint conditions of the strategy optimization model according to the initial operation strategy includes: Obtain the optimization objective information corresponding to the initial operation strategy, and determine the objective function of the strategy optimization model according to the optimization objective information; Obtain the strategy proportion information of each of the to-be-recommended user groups in the initial operation strategy, and determine the decision variables of the strategy optimization model according to the strategy proportion information; Obtain the operation resource parameters corresponding to the initial operation strategy, and determine the constraint conditions of the strategy optimization model according to the operation resource parameters.

8. A user operation strategy recommendation device, characterized in that Include: A user classification module, configured to obtain the target operation index and the user attribute data of all the to-be-recommended users, and perform classification processing on all the to-be-recommended users according to the target operation index and the user attribute data to obtain multiple to-be-recommended user groups; A strategy recommendation module, configured to perform data processing on user attribute data of all recommended users within a to-be-recommended user group through a preset strategy recommendation model corresponding to each to-be-recommended user group, so as to obtain an initial operation strategy corresponding to the to-be-recommended user group; An optimization model establishment module, configured to determine an objective function, decision variables, and constraint conditions of a strategy optimization model according to the initial operation strategy; An optimization model solution module, configured to perform solution processing on the strategy optimization model based on the objective function, decision variables, and constraint conditions, and determine an optimized operation strategy for each to-be-recommended user group according to the solution processing result; A target strategy determination module, configured to determine target operation strategy parameters of all to-be-recommended users according to the optimized operation strategy; 9. A computer device, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the computer-readable instructions, the user operation strategy recommendation method according to any one of claims 1 to 7 is implemented; 10. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the user operation strategy recommendation method according to any one of claims 1 to 7.

Images