Model training method, product recommendation method, equipment, medium and program product

Abstract

The invention discloses a model training method, a product recommendation method, equipment, a medium and a program product, and relates to the technical field of big data, and the model training method comprises the steps: obtaining initial feature information of a training user group; performing feature extraction on the initial feature information according to the correlation between each feature information in the initial feature information and the target product acceptance degree to obtain training feature information of the training user group; and training the to-be-trained model according to the training feature information and the target hyper-parameter group to obtain a recommendation group determination model. Therefore, on one hand, the correlation between the feature information and the target product acceptance degree is fully considered, and the group determination accuracy of the recommendation group determination model is improved. And on the other hand, only the target product is recommended to the user group, so that the user with relatively low acceptance on the target product is prevented from receiving the recommendation of the target product, and the marketing effect is improved. And on the other hand, the time cost in the marketing process is reduced.

Description

Technical Field

[0001] This application relates to the field of big data technology, and in particular to a model training method, a product recommendation method, equipment, media, and program products. Background Technology

[0002] Banks have a very large customer base, so when marketing wealth management products, it is necessary to identify potential customers within the bank's customer base and recommend the products to them.

[0003] Currently, the main marketing strategies employed at branch outlets include offering preferential deposit interest rates or reward points to attract new customers to purchase wealth management products. Additionally, they attract mobile banking users by sharing financial knowledge and investment tips on social media platforms.

[0004] However, both branch marketing and mobile banking content dissemination are marketing efforts aimed at all customers. Due to the large differences among customer groups and the varying levels of acceptance of financial products among each customer, some customers who do not need financial products receive information about them, resulting in poor marketing effectiveness and a significant waste of time in the marketing process. Summary of the Invention

[0005] This application provides a model training method, product recommendation method, device, medium, and program product to solve the problem in the prior art that due to the large differences in customer groups and the different acceptance levels of each customer towards financial products, some customers who do not need financial products receive information about financial products, resulting in poor marketing results and a waste of a lot of time in the marketing process.

[0006] Firstly, this application provides a model training method, the method comprising:

[0007] Obtain initial feature information of the training user group; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group;

[0008] Based on the correlation between each feature in the initial feature information and the acceptance level of the target product, feature extraction is performed on the initial feature information to obtain the training feature information of the training user group; wherein, the training feature information includes at least one feature information of the plurality of training users;

[0009] The training model is trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein, the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

[0010] Secondly, this application also provides a product recommendation method, the method comprising:

[0011] Obtain target feature information of the target user group; wherein, the target feature information includes at least one feature information of the same type as the training feature information of the model training method described in the first aspect of this application;

[0012] The target feature information of the target user group is input into the recommendation group determination model to obtain the recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is the model trained by the model training method described in the first aspect of this application;

[0013] The target product is pushed to the target recommended user group.

[0014] Thirdly, this application provides a model training apparatus, comprising:

[0015] The training acquisition module is used to acquire initial feature information of the training user group; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group;

[0016] The feature extraction module is used to extract features from the initial feature information based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, so as to obtain the training feature information of the training user group; wherein, the training feature information includes at least one feature information of the plurality of training users;

[0017] The training module is used to train the model to be trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein, the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

[0018] Fourthly, this application provides a product recommendation device, comprising:

[0019] The feature acquisition module is used to acquire target feature information of the target user group; wherein, the target feature information includes at least one type of information that is the same as the type of information of at least one feature information included in the training feature information of the model training method described in the first aspect of this application.

[0020] The group determination module is used to input the target feature information of the target user group into the recommendation group determination model to obtain the recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is the model trained by the model training method described in the first aspect of this application;

[0021] The push module is used to push the target product to the target recommended user group.

[0022] Fifthly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the model training method as described in the first aspect of this application, or the product recommendation method as described in the second aspect of this application.

[0023] Sixthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model training method as described in the first aspect of this application, or the product recommendation method as described in the second aspect of this application.

[0024] In a seventh aspect, this application provides a computer program product, including a computer program that, when executed by a processor, implements the model training method as described in the first aspect of this application, or the product recommendation method as described in the second aspect of this application.

[0025] The model training method of this application obtains initial feature information of a training user group; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group; based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, feature extraction is performed on the initial feature information to obtain training feature information of the training user group; wherein, the training feature information includes at least one feature information of multiple training users; the model to be trained is trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein, the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product. That is, the model training method of this application, on the one hand, performs feature extraction on the initial feature information based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, fully considering the correlation between feature information and the acceptance level of the target product, thereby improving the accuracy of the recommendation group determination model in determining the user group. On the other hand, by identifying the target user group through the recommendation group determination model, it is possible to pinpoint the user group with a high acceptance level of the target product. This allows the target product to be recommended only to this user group, avoiding recommendations to users with low acceptance levels and improving marketing effectiveness. Furthermore, the model's fast processing speed and high efficiency mean that using it to determine the user group can reduce time costs in the marketing process. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating the model training method provided in this application;

[0028] Figure 2 This is another flowchart illustrating the model training method provided in this application;

[0029] Figure 3 This is a flowchart illustrating the product recommendation method provided in this application;

[0030] Figure 4 This is a schematic diagram of the model training device provided in this application;

[0031] Figure 5 This is a schematic diagram of the product recommendation device provided in this application;

[0032] Figure 6 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0033] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0034] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0035] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0036] Figure 1 This is a flowchart illustrating the model training method provided in this application. This method can be executed by the model training device provided in this application, which can be implemented in software and / or hardware. In a specific embodiment, the device can be integrated into an electronic device. For example, the device can be integrated into a computer. The following embodiments will illustrate this using the integration of the device into an electronic device as an example. Reference Figure 1 The method may specifically include the following steps:

[0037] Step 101: Obtain the initial feature information of the training user group.

[0038] The initial feature information includes various feature information of multiple training users in the training user group.

[0039] Specifically, the training user group can be a portion of historical user groups uploaded after staff have compiled statistical information, thereby increasing the accuracy of the training data. The statistical information constitutes the initial characteristic information of the training user group. Therefore, the initial characteristic information includes multiple characteristic information of multiple training users within the training user group, with each training user possessing multiple characteristic information. For example, the sources of training data collection can include application statistics, staff statistics, and public network data analysis. The initial characteristic information is partial information that can characterize user traits. For example, the initial characteristic information can include the user's personal identification information, the user's age group, the user's occupation, the user's economic level, and whether the user has actively inquired about the target product. The target product is the recommended product in this embodiment. For example, the target product can be a financial product.

[0040] Step 102: Based on the correlation between each feature in the initial feature information and the acceptance level of the target product, feature extraction is performed on the initial feature information to obtain the training feature information of the training user group.

[0041] The training feature information includes at least one feature information from multiple training users.

[0042] Specifically, a user's acceptance level of the target product is used to characterize the suitability of recommending the target product to the user. For example, if a user has a high acceptance level of the target product, then the target product is deemed suitable for recommendation. In the initial feature information, some data may be unrelated or only weakly correlated with the target product acceptance level. Therefore, it is necessary to analyze and calculate the correlation between the two variables. For example, a user's personal identification information is unrelated to their target product acceptance level, thus determining that there is no correlation between the two. However, whether a user has actively consulted about the target product is related to their target product acceptance level, thus determining that there is a high correlation between the two. Therefore, feature extraction is performed on the features in the initial feature information that have a correlation greater than a preset value with the user's target product acceptance level to obtain the training feature information for the training user group. Feature extraction refers to dimensionality reduction based on the selected features. Reducing the dimensionality of the data reduces computational requirements, and this can be achieved through kernel principal component analysis (KPCA).

[0043] For example, five characteristics of the user group were selected, such as the user's personal identification information, age group, occupation, economic level, and whether the user has actively consulted the target product. By using the kernel principal component analysis method for feature extraction, the characteristics of the user group can be reduced to three, namely the user's occupation, economic level, and whether the user has actively consulted the target product, thereby reducing the model complexity for subsequent calculations.

[0044] Optionally, after obtaining the initial feature information, the initial feature information can also be cleaned. Data cleaning may include filling or deleting missing data to handle missing values, and identifying and processing outliers to detect outliers, thereby cleaning the data and making the data in a uniform format, which is convenient for subsequent data processing.

[0045] Optionally, steps 21 to 22 may be performed before step 103.

[0046] Step 21: Obtain the initial hyperparameter set of the model to be trained.

[0047] The initial hyperparameter set includes multiple hyperparameters.

[0048] Specifically, hyperparameters are parameters in machine learning and deep learning models that affect the model's performance, such as learning rate, batch size, number of iterations, regularization parameters, and activation functions. Based on historical data, an initial set of hyperparameters for the model to be trained is obtained.

[0049] Step 22: Optimize the initial hyperparameter set based on the whale optimization algorithm to determine the optimal hyperparameter set as the target hyperparameter set.

[0050] Specifically, by optimizing the initial hyperparameter set using the whale optimization algorithm, the final optimal solution is obtained by iterating through multiple optimal solutions of the hyperparameter set. This final optimal solution is then taken as the target hyperparameter set. Obtaining the optimal hyperparameter set using the whale optimization algorithm makes the determined target hyperparameter set more conducive to the model training process, thereby shortening the model training time and improving the model training efficiency.

[0051] Optionally, step 22 can be implemented through steps 221 to 225.

[0052] Step 221: Randomly generate a whale population based on the initial hyperparameter set.

[0053] In this context, individual whales in the whale population are used to represent multiple solutions to the optimization process of the initial hyperparameter set. The whale population includes multiple first whale individuals and one optimal whale individual.

[0054] Specifically, in the whale optimization algorithm, a whale population is randomly generated based on the initial hyperparameter set. Each individual whale in the whale population represents the solution of the optimization process of the initial hyperparameter set. In the initially randomly generated whale population, the whale population includes multiple first whale individuals and one optimal whale individual.

[0055] Optionally, a whale population can be randomly generated based on the initial hyperparameter set and the chaotic mapping algorithm.

[0056] Specifically, a whale population can be randomly generated using a chaotic mapping algorithm. Using chaotic mapping instead of initializing the whale population avoids the uneven distribution of whales (too close or too far apart) caused by pseudo-random numbers during initialization, which compromises population diversity. Chaotic mapping, with its regularity, randomness, and ergodicity, enhances population diversity. Chaotic mapping can be implemented using Equation 1.

[0057] λ t+1 =μ×λ t (1-λ t ), t=0,1,2...,T Formula 1

[0058] Where, λ t+1 Let λ represent the population proportion in the (t+1)th iteration. t Let represent the population proportion in the t-th iteration, which is any random number uniformly distributed on [0,1]. t represents the iteration number, T represents the maximum number of chaotic iterations, and μ represents the chaos control parameter, which can be 4. The whale population can be generated using Equation 2 based on the chaotic mapping algorithm.

[0059]

[0060] in, Let λ represent the coordinates of the j-th dimension in the i-th search agent group. j Let λ represent the coordinates of the j-th dimension of λ after internal random sorting, lb be the lower bound of the search space, and ub be the upper bound of the search space.

[0061] Step 222: Based on the distance between each first whale individual and the optimal whale individual and the adaptive weight strategy algorithm, update each first whale individual to obtain multiple second whale individuals.

[0062] Specifically, based on the distance between each first whale individual and the optimal whale individual, an adaptive weight strategy algorithm is used to move each first whale individual in the direction closest to the optimal whale individual. That is, different distances are moved according to different weights. Then, each first whale individual is updated to obtain multiple new first whale individuals, which in turn result in multiple second whale individuals. The introduction of the adaptive weight algorithm can automatically adjust the weight values ​​according to the iteration process of the algorithm, thereby finding the global optimum more efficiently while maintaining a fast convergence speed.

[0063] Optionally, each first whale individual is updated based on the distance between each first whale individual and the optimal whale individual, an adaptive weighting strategy algorithm, and a nonlinear decreasing balance algorithm to obtain multiple second whale individuals.

[0064] Specifically, based on the distance between each first whale individual and the optimal whale individual, and the adaptive weight strategy algorithm, each first whale individual is updated to obtain multiple second whale individuals. During the update, a nonlinear decreasing balancing algorithm is used to update the coefficient vector. In the whale optimization algorithm, the coefficient vector is used to determine the change in the whale individual's position. Therefore, by introducing a nonlinear decreasing balancing algorithm to change the coefficient vector, the distance updated by the whale individual can be changed. Thus, the nonlinear decreasing balancing algorithm has high global and local search capabilities, which can improve the convergence accuracy of the algorithm. Changing the coefficient vector of the whale optimization algorithm using a nonlinear decreasing balancing algorithm can be achieved through Equation 3.

[0065]

[0066] Where t represents the number of iterations, T represents the maximum number of iterations, m and n are user-defined parameters, and the values ​​of m and n are set to 2 and 6 respectively. a(t) represents the coefficient vector at the t-th iteration.

[0067] Step 223: If the fitness of multiple second whale individuals is less than or equal to the fitness of the optimal whale individual, then the multiple second whale individuals are treated as multiple new first whale individuals, and the process returns to step 222 until the number of iterations is greater than or equal to the preset number.

[0068] Specifically, the fitness of a whale individual is used to characterize the degree of optimization of the solution corresponding to that whale individual. If the fitness of multiple second whale individuals is less than or equal to the fitness of the optimal whale individual, then the optimal solution corresponding to the optimal whale individual is still the optimal solution in the current iteration process. Therefore, the optimal solution is not changed at this time, and the second whale individual is taken as the new first whale individual, resulting in multiple first whale individuals. The process returns to step 222 until the number of iterations is greater than or equal to the preset number. The preset number is a preset number of iterations that can determine the convergence of the whale optimization algorithm. For example, the preset number can be 1000 iterations.

[0069] Step 224: For each second whale individual, if the fitness of the second whale individual is greater than the fitness of the optimal whale individual, then the second whale individual is taken as the new optimal whale individual, and the optimal whale individual and the second whale individuals other than the second whale individual among the multiple second whale individuals are taken as the new multiple first whale individuals. Then return to step 222 and continue until the number of iterations is greater than or equal to the preset number of iterations.

[0070] Specifically, for each second whale individual, if the fitness of the second whale individual is greater than the fitness of the optimal whale individual, then the optimal solution corresponding to the second whale individual is the optimal solution in the current iteration process. At this time, the second whale individual is taken as the new optimal whale individual, and the optimal whale individual and the second whale individuals other than the second whale individual among the multiple second whale individuals are taken as the new multiple first whale individuals, and the process returns to step 222 until the number of iterations is greater than or equal to the preset number.

[0071] Step 225: Use the hyperparameter set corresponding to the optimal whale individual as the target hyperparameter set.

[0072] Specifically, when the number of iterations is greater than or equal to the preset number, it indicates that the whale optimization algorithm has converged sufficiently. At this point, the hyperparameter set corresponding to the optimal whale individual is taken as the target hyperparameter set.

[0073] Step 103: Train the model to be trained based on the training feature information and the target hyperparameter set to obtain the recommendation group determination model.

[0074] The recommendation group determination model is used to output the recommendation group determination result based on the target characteristic information of the target user group. The recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

[0075] Specifically, the training feature information is input into the training model (with the target hyperparameter set as the input hyperparameter set) for training. The model parameters are adjusted based on the training loss to obtain the recommendation group determination model. For example, training can be stopped when a certain cutoff condition is reached, and the model updated in the last round is used as the recommendation group determination model. The cutoff condition could be reaching a preset number of training rounds, or the training loss falling below a certain value. The training model can be a Long Short-Term Memory (LSTM) neural network model. The recommendation group determination model can output recommendation group determination results based on the target feature information of the target user group. The recommendation group determination results are used to indicate the target recommended user group corresponding to the target product.

[0076] The model training method of this application obtains initial feature information of a training user group; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group; based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, feature extraction is performed on the initial feature information to obtain training feature information of the training user group; wherein, the training feature information includes at least one feature information of multiple training users; the model to be trained is trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein, the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product. That is, the model training method of this application, on the one hand, performs feature extraction on the initial feature information based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, fully considering the correlation between feature information and the acceptance level of the target product, thereby improving the accuracy of the recommendation group determination model in determining the user group. On the other hand, by identifying the target user group through the recommendation group determination model, it is possible to pinpoint the user group with a high acceptance level of the target product. This allows the target product to be recommended only to this user group, avoiding recommendations to users with low acceptance levels and improving marketing effectiveness. Furthermore, the model's fast processing speed and high efficiency mean that using it to determine the user group can reduce time costs in the marketing process.

[0077] Figure 2 This is another flowchart illustrating the model training method provided in this application. This embodiment... Figure 1 Based on the illustrated embodiments and various optional implementation schemes, the steps for determining the training feature information of the training user group are described in detail. For example... Figure 2 As shown, the method may include the following steps:

[0078] Step 201: Obtain the initial feature information of the training user group.

[0079] Step 202: For each type of feature information, calculate the product-moment correlation coefficient and the rank correlation coefficient between the feature information and the acceptance level of the target product.

[0080] Specifically, the product-moment correlation coefficient (PMC) is a linear correlation coefficient used to measure the strength and direction of the linear relationship between two variables. A PMC greater than 0 indicates a positive correlation between the two variables. The rank correlation coefficient is a statistic used to assess the strength and direction of the monotonic relationship between two variables. When the rank correlation coefficient is positive, it means that as one variable increases, the other variable also tends to increase. Therefore, by calculating the PMC and rank correlation coefficients of each feature information with the acceptance level of the target product, the correlation between each feature information and the acceptance level of the target product can be determined.

[0081] Step 203: When the product-moment correlation coefficient is greater than the first preset value and the rank correlation coefficient is greater than the second preset value, it is determined that the correlation between the feature information and the acceptance level of the target product meets the preset correlation requirements.

[0082] Specifically, the first preset value is a preset product-moment correlation coefficient that represents the correlation between the feature information and the acceptance level of the target product. For example, the first preset value can be 0. The second preset value is a preset rank correlation coefficient that represents the correlation between the feature information and the acceptance level of the target product. For example, the second preset value can be 0.5. The first and second preset values ​​can be the same or different, and this embodiment does not limit this. The preset correlation requirement is a preset requirement that the correlation between the feature information and the acceptance level of the target product is higher than a preset correlation requirement. For example, the preset correlation requirement is that the product-moment correlation coefficient is greater than 0.2 and the rank correlation coefficient is greater than 0.5. Therefore, when the product-moment correlation coefficient is greater than the first preset value and the rank correlation coefficient is greater than the second preset value, it can be determined that the correlation between the feature information and the acceptance level of the target product meets the preset correlation requirement.

[0083] Step 204: Extract features from the initial feature information that meet the preset relevant requirements to obtain the training feature information of the training user group.

[0084] Specifically, feature extraction is performed on the initial feature information that meets the preset relevance requirements. That is, feature extraction is performed on the feature information whose correlation with the target product's acceptance level is higher than the preset relevance, so as to obtain the training feature information of the training user group.

[0085] Step 205: Train the model to be trained based on the training feature information and the target hyperparameter set to obtain the recommendation group determination model.

[0086] The model training method in this application determines the correlation between feature information and the acceptance level of the target product by calculating the product-moment correlation coefficient and the rank correlation coefficient between feature information and the acceptance level of the target product. This further avoids recommending the target product to users with low acceptance of the target product, thereby improving the accuracy of the group determination model for the recommendation group determination.

[0087] Figure 3 This is a flowchart illustrating a product recommendation method provided in this application. This method can be executed by a product recommendation device provided in this application, which can be implemented using software and / or hardware. In a specific embodiment, the device can be integrated into an electronic device. For example, the device can be integrated into a computer. The following embodiments will be described using the integration of the device into an electronic device as an example. (Reference) Figure 3 The method may specifically include the following steps:

[0088] Step 301: Obtain target characteristic information of the target user group.

[0089] The target feature information includes at least one feature information of the same type as the training feature information of the model training method in any embodiment of this application.

[0090] Specifically, the target feature information includes at least one type of information that is the same as the type of information included in the training feature information of the model training method in any embodiment of this application. Therefore, the feature information in the target feature information is highly correlated with the acceptance level of the target product, thereby increasing the accuracy of the model calculation results. The target user group is the user group whose need to be determined for recommending the target product is to be determined. This group may include three types of people, such as users who have not yet purchased the target product, users who have the purchasing power but no intention to purchase the target product, and users who have both the purchasing power and the intention to purchase the target product. Therefore, it is necessary to obtain the target feature information of the target user group.

[0091] Step 302: Input the target feature information of the target user group into the recommendation group determination model to obtain the recommendation group determination result corresponding to the target user group.

[0092] The recommended user group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommended user group determination model is the model trained by the model training method of any embodiment of this application.

[0093] Specifically, after obtaining the target feature information of the target user group, the target feature information of the target user group is input into the recommendation group determination model, and the output of the model is the recommendation group determination result corresponding to the target user group.

[0094] Step 303: Push the target product to the target recommended user group.

[0095] Specifically, after obtaining the target user group, the target product is pushed to this group. Recommendation methods can include online push notifications to each user's device within the target user group, or sending the target user group information to staff to contact and recommend the product to users within that group. After pushing the target product to the target user group, the model can be evaluated based on the actual push results. For example, the actual results can be defined as the target user group's acceptance of the target product and the number of visits to related target products. For instance, before determining the target user group, staff might use phone calls to recommend the product to users who express a willingness to purchase. After determining the target user group, staff might use phone calls again to recommend the product to users who express a willingness to purchase. Both methods involve making the same number of calls, and the ratio of users with a purchase intention to the number of calls made is calculated; this ratio is used as an evaluation metric for the model's performance. If the increase in the evaluation metric is greater than the preset range, it indicates that the model is performing well. If the increase in the evaluation metric is less than or equal to the preset range, it indicates that the model's performance in identifying the target recommended audience has not met expectations. In this case, it is necessary to continue optimizing the model. Model optimization can be achieved by adjusting the model's hyperparameters.

[0096] The product recommendation method of this application obtains target feature information of a target user group; wherein, the target feature information includes at least one type of information that is the same as the type of information of at least one type of information of the training feature information of the model training method described in the first aspect of this application; the target feature information of the target user group is input into a recommendation group determination model to obtain a recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is a model trained by the model training method of any embodiment of this application; the target product is then pushed to the target recommended user group. That is, the product recommendation method of this application, on the one hand, ensures that the type of information of the target feature information is the same as the type of information of the training feature information of the model training method, which enables a high correlation between the feature information of the target feature information and the acceptance level of the target product, thereby improving the accuracy of the recommendation group determination result. On the other hand, the recommendation group determination result obtained by determining the recommendation group based on the recommendation group determination model can be used to determine the target recommended user group and recommend the target product to the target recommended user group, thereby recommending the target product only to user groups with a high acceptance level of the target product, thus avoiding recommendations to users with a low acceptance level of the target product and improving marketing effectiveness.

[0097] Figure 4This is a schematic diagram of the model training apparatus provided in this application, which is suitable for executing the model training method provided in this application. Figure 4 As shown, the device may specifically include:

[0098] The training acquisition module 401 is used to acquire initial feature information of the training user group; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group.

[0099] The feature extraction module 402 is used to extract features from the initial feature information based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, so as to obtain the training feature information of the training user group; wherein, the training feature information includes at least one feature information of the plurality of training users.

[0100] The training module 403 is used to train the model to be trained based on the training feature information and the target hyperparameter group to obtain the recommendation group determination model; wherein, the recommendation group determination model is used to output the recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

[0101] In one embodiment, the feature extraction module 402 is specifically used to: calculate the product-moment correlation coefficient and the rank correlation coefficient between the feature information and the acceptance level of the target product for each type of feature information; when the product-moment correlation coefficient is greater than a first preset value and the rank correlation coefficient is greater than a second preset value, determine that the correlation between the feature information and the acceptance level of the target product meets a preset correlation requirement; and extract features from the initial feature information that meet the preset correlation requirement to obtain the training feature information of the training user group.

[0102] In one embodiment, the apparatus further includes: a hyperparameter set determination module, configured to obtain an initial hyperparameter set of the model to be trained before the training module 403 trains the model to be trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein the initial hyperparameter set includes multiple hyperparameters; and an optimization module, configured to optimize the initial hyperparameter set based on the whale optimization algorithm to determine the optimal hyperparameter set as the target hyperparameter set.

[0103] In one embodiment, the optimization module is specifically used to: randomly generate a whale population based on the initial hyperparameter set; wherein, individual whales in the whale population are used to represent multiple solutions of the optimization process of the initial hyperparameter set, and the whale population includes multiple first whales and one optimal whale; update each first whale based on the distance between each first whale and the optimal whale and an adaptive weight strategy algorithm to obtain multiple second whales; if the fitness of each of the multiple second whales is less than or equal to the fitness of the optimal whale, then the multiple second whales are used as new multiple first whales, and the process returns to execute "updating each first whale based on the distance between each first whale and the optimal whale and an adaptive weight strategy algorithm to obtain multiple solutions of the optimization process of the initial hyperparameter set, and the optimization process of the initial hyperparameter set includes multiple first whales and one optimal whale; update each first whale based on the distance between each first whale and the optimal whale and an adaptive weight strategy algorithm to obtain multiple solutions of the optimization process of the initial hyperparameter set, and obtain multiple solutions of the optimization process of the initial hyperparameter set, and the optimization process of the optimization process of the initial hyperparameter set includes ... The process of "updating individual whales to obtain multiple second whales" continues until the number of iterations is greater than or equal to a preset number. For each second whale, if the fitness of the second whale is greater than the fitness of the optimal whale, then the second whale is taken as the new optimal whale. The optimal whale and the second whales other than the second whale among the multiple second whales are taken as new multiple first whales. The process of "updating each first whale based on the distance between each first whale and the optimal whale and the adaptive weight strategy algorithm to obtain multiple second whales" continues until the number of iterations is greater than or equal to a preset number. The hyperparameter set corresponding to the optimal whale is taken as the target hyperparameter set.

[0104] In one embodiment, the optimization module, in the process of randomly generating a whale population based on the initial hyperparameter set, is specifically used to: randomly generate a whale population based on the initial hyperparameter set and a chaotic mapping algorithm.

[0105] In one embodiment, the optimization module updates each first whale individual based on the distance between each first whale individual and the optimal whale individual and an adaptive weight strategy algorithm to obtain multiple second whale individuals. Specifically, this is done by: updating each first whale individual based on the distance between each first whale individual and the optimal whale individual, the adaptive weight strategy algorithm, and a nonlinear decreasing balance algorithm to obtain multiple second whale individuals.

[0106] The model training apparatus of this application has the same optional implementation methods and the beneficial effects that can be achieved as those in the above-described model training method embodiments, and will not be repeated here.

[0107] Figure 5 This is a schematic diagram of a product recommendation device provided in this application, which is suitable for executing the product recommendation method provided in this application. Figure 5As shown, the device may specifically include:

[0108] The feature acquisition module 501 is used to acquire target feature information of the target user group; wherein, the target feature information includes at least one type of feature information that is the same as the type of feature information included in the training feature information of the model training method described in any embodiment of this application.

[0109] The group determination module 502 is used to input the target feature information of the target user group into the recommendation group determination model to obtain the recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is the model trained by the model training method described in any embodiment of this application.

[0110] The push module 503 is used to push the target product to the target recommended user group.

[0111] The product recommendation device of this application has the same optional implementation methods and the same beneficial effects as the product recommendation method embodiments described above, and will not be repeated here.

[0112] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the model training method provided in any of the above embodiments, or the product recommendation method provided in any of the above embodiments.

[0113] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model training method provided in any of the above embodiments, or implements the product recommendation method provided in any of the above embodiments.

[0114] The following is for reference. Figure 6 It shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the present application. Figure 6 The electronic device shown is merely an example and should not impose any limitations on the functionality and scope of this application.

[0115] like Figure 6As shown, the electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the electronic device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0116] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0117] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs the functions defined above in the system of this application.

[0118] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0119] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of the boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0120] The modules and / or units described in this application can be implemented in software or hardware. The described modules and / or units can also be housed in a processor; for example, a processor may be described as including a training acquisition module, a feature extraction module, and a training module. Alternatively, a processor may be described as including a feature acquisition module, a population determination module, and a push module. The names of these modules do not necessarily constitute a limitation on the module itself.

[0121] In another aspect, this application also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist alone and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include:

[0122] The process involves: acquiring initial feature information of the training user group; where the initial feature information includes multiple feature information of multiple training users in the training user group; extracting features from the initial feature information based on the correlation between each feature information and the acceptance level of the target product, thereby obtaining training feature information of the training user group; where the training feature information includes at least one feature information of multiple training users; training the model to be trained based on the training feature information and the target hyperparameter set, thereby obtaining a recommendation group determination model; where the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

[0123] According to the model training method of this application, initial feature information of the training user group is obtained; wherein, the initial feature information includes multiple feature information of multiple training users in the training user group; based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, feature extraction is performed on the initial feature information to obtain training feature information of the training user group; wherein, the training feature information includes at least one feature information of multiple training users; the model to be trained is trained based on the training feature information and the target hyperparameter set to obtain a recommendation group determination model; wherein, the recommendation group determination model is used to output a recommendation group determination result based on the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product. That is, the model training method of this application, on the one hand, performs feature extraction on the initial feature information based on the correlation between each feature information in the initial feature information and the acceptance level of the target product, fully considering the correlation between feature information and the acceptance level of the target product, thereby improving the accuracy of the recommendation group determination model in determining the user group. On the other hand, by identifying the target user group through the recommendation group determination model, it is possible to pinpoint the user group with a high acceptance level of the target product. This allows the target product to be recommended only to this user group, avoiding recommendations to users with low acceptance levels and improving marketing effectiveness. Furthermore, the model's fast processing speed and high efficiency mean that using it to determine the user group can reduce time costs in the marketing process.

[0124] Alternatively, when one or more of the above programs are executed by the device, the device includes:

[0125] Obtain target feature information of the target user group; wherein, the target feature information includes at least one feature information of the same type as the training feature information of the model training method described in the first aspect of this application; input the target feature information of the target user group into the recommendation group determination model to obtain the recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is the model trained by the model training method of any embodiment of this application; push the target product to the target recommended user group.

[0126] According to the product recommendation method of this application, target feature information of a target user group is obtained; wherein, the information type of at least one feature information included in the target feature information is the same as the information type of at least one feature information included in the training feature information of the model training method described in the first aspect of this application; the target feature information of the target user group is input into a recommendation group determination model to obtain a recommendation group determination result corresponding to the target user group; wherein, the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is a model trained by the model training method of any embodiment of this application; the target product is pushed to the target recommended user group. That is, the product recommendation method of this application, on the one hand, the information type of the feature information included in the target feature information is the same as the information type of the feature information included in the training feature information of the model training method, which can make the feature information included in the target feature information highly correlated with the acceptance level of the target product, thereby improving the accuracy of the recommendation group determination result. On the other hand, the recommendation group determination result obtained by determining the recommendation group based on the recommendation group determination model can be used to determine the target recommended user group and recommend the target product to the target recommended user group, thereby recommending the target product only to the user group with a high acceptance level of the target product, thus avoiding the target product being recommended to users with a low acceptance level of the target product, and improving the marketing effect.

[0127] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the model training method provided in any embodiment of this application, or implements the product recommendation method provided in any embodiment of this application.

[0128] In the implementation of the computer program product, computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0129] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0130] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A model training method, characterized in that: include: Acquire initial feature information of a training user group; wherein the initial feature information includes multiple feature information of multiple training users in the training user group; According to the correlation between each type of feature information in the initial feature information and the acceptance degree of the target product, feature extraction is performed on the initial feature information to obtain training feature information of the training user group; wherein the training feature information includes at least one type of feature information of the multiple training users; The model to be trained is trained according to the training feature information and the target hyperparameter group to obtain a recommendation group determination model; wherein the recommendation group determination model is used to output a recommendation group determination result according to the target feature information of the target user group, and the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product.

2. The method according to claim 1, characterized in that: The extracting features of the initial feature information according to the correlation between each feature information in the initial feature information and the acceptance degree of the target product to obtain the training feature information of the training user group includes: For each type of feature information, calculating the product-moment correlation coefficient and the rank correlation coefficient between the feature information and the acceptance degree of the target product; When the product-moment correlation coefficient is greater than a first preset value, and the rank correlation coefficient is greater than a second preset value, it is determined that the correlation between the feature information and the acceptance degree of the target product meets a preset correlation requirement; Feature extraction is performed on the feature information that meets the preset related requirements in the initial feature information to obtain training feature information of the training user group.

3. The method according to claim 1, characterized in that Before training the to-be-trained model according to the training feature information and the target hyperparameter group to obtain the recommended group determination model, the method further includes: Obtaining an initial hyperparameter group of the model to be trained; wherein the initial hyperparameter group includes multiple hyperparameters; The initial hyperparameter group is optimized based on the whale optimization algorithm to determine the optimal hyperparameter group as the target hyperparameter group.

4. The method according to claim 3, characterized in that: The step of optimizing the initial hyperparameter group based on the whale optimization algorithm and determining the optimal hyperparameter group as the target hyperparameter group includes: A whale population is randomly generated according to the initial hyperparameter group; wherein the whale individuals in the whale population are used to characterize multiple solutions of the optimization process of the initial hyperparameter group, and the whale population includes multiple first whale individuals and an optimal whale individual; According to the distance between each of the first whale individuals and the optimal whale individual and the adaptive weight strategy algorithm, each of the first whale individuals is updated to obtain a plurality of second whale individuals; If the fitness corresponding to the plurality of second whale individuals is less than or equal to the fitness corresponding to the optimal whale individual, the plurality of second whale individuals are taken as new plurality of first whale individuals, and the process returns to execute the step of "updating each first whale individual according to the distance between each first whale individual and the optimal whale individual and the adaptive weight strategy algorithm to obtain a plurality of second whale individuals" until the number of iterations is greater than or equal to the preset number; For each of the second whale individuals, if the fitness corresponding to the second whale individual is greater than the fitness corresponding to the optimal whale individual, the second whale individual is used as a new optimal whale individual, and the optimal whale individual and the second whale individuals other than the second whale individual among the multiple second whale individuals are used as new multiple first whale individuals, and the step of "updating each of the first whale individuals according to the distance between each of the first whale individuals and the optimal whale individual and the adaptive weight strategy algorithm to obtain multiple second whale individuals" is returned to be executed until the number of iterations is greater than or equal to the preset number; The hyperparameter group corresponding to the optimal whale individual is used as the target hyperparameter group.

5. The method according to claim 4, characterized in that The randomly generating a whale population according to the initial hyperparameter group comprises: According to the initial hyperparameter group and the chaotic mapping algorithm, a whale population is randomly generated.

6. The method according to claim 4, characterized in that The method updates each of the first whale individuals according to the distance between each of the first whale individuals and the optimal whale individual and the adaptive weight strategy algorithm to obtain multiple second whale individuals, including: According to the distance between each of the first whale individuals and the optimal whale individual, the adaptive weight strategy algorithm and the nonlinear decreasing equilibrium algorithm, each of the first whale individuals is updated to obtain a plurality of second whale individuals.

7. A product recommendation method, characterized in that: include: Obtain target feature information of a target user group; wherein the target feature information includes at least one feature information of the same type as the at least one feature information included in the training feature information of the model training method according to any one of claims 1 to 6; Inputting the target feature information of the target user group into a recommendation group determination model to obtain a recommendation group determination result corresponding to the target user group; wherein the recommendation group determination result is used to indicate the target recommended user group corresponding to the target product, and the recommendation group determination model is a model trained according to the model training method described in any one of claims 1 to 6; The target product is pushed to the target recommendation user group.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the model training method as described in any one of claims 1 to 6, or, when executing the program, it implements the product recommendation method as described in claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the model training method as described in any one of claims 1 to 6, or, when the program is executed, it implements the product recommendation method as described in claim 7.

10. A computer program product, comprising a computer program, characterized in that When executed by a processor, the computer program implements the model training method as described in any one of claims 1 to 6, or implements the product recommendation method as described in claim 7.

Images