Product recommendation method and device based on marketing model, equipment and storage medium

By processing and screening data samples multiple times and using the LCE algorithm to build a marketing model, the problems of insufficient model stability and high-dimensional data processing in the existing technology are solved, and the accuracy of customer group screening and product marketing effect are improved.

CN119066523AInactive Publication Date: 2024-12-03NANJING XINGLIAN DIGITAL TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411562440.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-05
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing marketing model based on logistic regression has problems such as multicollinearity sensitivity of independent variables when processing big data, difficulty in dealing with high-dimensional variables and insufficient model stability, resulting in poor customer group screening accuracy.

Method used

By processing and screening data samples multiple times, using the LCE algorithm to build a marketing model, and training and verification of the built models, the model prediction and distinction ability and stability are improved, thereby improving the accuracy of customer group screening.

Benefits of technology

It improves the accuracy of target user group screening and the effectiveness of product marketing, solves the problem of high-dimensional data processing, and improves the prediction and distinction capabilities of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066523A_ABST
    Figure CN119066523A_ABST
Patent Text Reader

Abstract

The invention relates to a product recommendation method and device based on a marketing model, equipment and a storage medium, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining and preprocessing a first sample data set based on a preset sample template, and obtaining a second sample data set; retaining corresponding data when the stability index value of the data in the second sample data set is smaller than or equal to a first preset threshold value to generate a third sample data set; training and verifying a first marketing model constructed based on an LCE algorithm by using the third sample data set, and outputting a second marketing model when a verification result meets a preset standard; and inputting the real-time customer information into the second marketing model to determine a target user group needing to recommend the product information. According to the method, the problem of high-dimensional data processing can be solved, the prediction and distinguishing capabilities of the model and the stability of the model can be improved, so that the accuracy of screening the target user group is improved, and the effectiveness of product marketing is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a product recommendation method, device, equipment and storage medium based on a marketing model. Background Art

[0002] The existing scoring card marketing model based on logistic regression (LR) has the advantages of strong interpretability, fast calculation speed, and simplicity and ease of understanding. However, with the development of big data, the limitations of the logistic regression algorithm have also become prominent. For example, it is sensitive to multicollinearity of independent variables in the model, feature engineering is difficult to handle variables with larger dimensions, and there is a lack of model stability.

[0003] Therefore, there is an urgent need to propose a product recommendation method, device, equipment and storage medium based on a marketing model that can improve the accuracy of customer group screening. Summary of the invention

[0004] Based on this, it is necessary to provide a product recommendation method, device, equipment and storage medium based on a marketing model that can improve the accuracy of customer group screening in response to the above technical problems.

[0005] In a first aspect, a product recommendation method based on a marketing model is provided, the method comprising: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

[0006] Optionally, based on a preset sample template, obtaining the first sample data set includes: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0007] Optionally, preprocessing the first sample data set to obtain a second sample data set includes: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0008] Optionally, before obtaining the stability index value of the data in the second sample data set, the method further includes: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: Based on this, it is necessary to provide a product recommendation method, device, equipment and storage medium based on a marketing model that can improve the accuracy of customer group screening in response to the above technical problems.

[0009] In a first aspect, a product recommendation method based on a marketing model is provided, the method comprising: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

[0010] Optionally, based on a preset sample template, obtaining the first sample data set includes: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0011] Optionally, preprocessing the first sample data set to obtain a second sample data set includes: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0012] Optionally, before obtaining the stability index value of the data in the second sample data set, the method further includes: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: Based on this, it is necessary to provide a product recommendation method, device, equipment and storage medium based on a marketing model that can improve the accuracy of customer group screening in response to the above technical problems.

[0013] In a first aspect, a product recommendation method based on a marketing model is provided, the method comprising: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

[0014] Optionally, based on a preset sample template, obtaining the first sample data set includes: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0015] Optionally, preprocessing the first sample data set to obtain a second sample data set includes: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0016] Optionally, before obtaining the stability index value of the data in the second sample data set, the method further includes: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity; In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin.

[0017] Optionally, obtaining a stability index value of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to a first preset threshold, to generate a third sample data set includes: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, retaining data corresponding to the stability index value; The retained data is sampled according to a preset ratio to generate a third sample data set.

[0018] Optionally, constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, in response to detecting that the verification result of the first marketing model meets a preset standard, outputting a second marketing model includes: Based on the LCE algorithm, an initial marketing model is constructed, and parameters of the initial marketing model are adjusted to obtain the first marketing model; Based on the training set and the validation set in the third sample data set, respectively training and validating the first marketing model; In response to detecting that the verification result meets the preset standard, a second marketing model is output.

[0019] Optionally, after obtaining the second marketing model, the method further includes: Obtaining a first evaluation index value and a second evaluation index value of the second marketing model; Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated.

[0020] In a second aspect, a product recommendation device based on a marketing model is provided, the device comprising: A first data processing module, used to obtain a first sample data set based on a preset sample template, and preprocess the first sample data set to obtain a second sample data set; A second data processing module is used to obtain the stability index value of the data in the second sample data set, and retain the data corresponding to when the stability index value of the data is less than or equal to the first preset threshold value, so as to generate a third sample data set; a marketing model generation module, configured to construct a first marketing model based on the LCE algorithm, and train and verify the first marketing model using the third sample data set, and output a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; The user group determination module is used to obtain real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

[0021] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

[0022] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

[0023] The above-mentioned product recommendation method, device, equipment and storage medium based on the marketing model, the method includes: based on a preset sample template, obtaining a first sample data set, and preprocessing the first sample data set to obtain a second sample data set; obtaining the stability index value of the data in the second sample data set, retaining the data corresponding to the stability index value of the data is less than or equal to the first preset threshold value to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, in response to detecting that the verification result of the first marketing model meets the preset standard, outputting the second marketing model; obtaining real-time customer information, and inputting it into the second marketing model to determine the target user group, and recommending product information to the target user group. The present application can solve the problem of high-dimensional data processing by processing and screening data samples multiple times, constructing a marketing model based on the LCE algorithm, and using the screened data to train and verify the constructed marketing model, which can improve the prediction and differentiation capabilities of the model and the stability of the model, so as to improve the accuracy of screening the target user group, thereby improving the effectiveness of product marketing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 An application environment diagram of a product recommendation method based on a marketing model in one embodiment; Figure 2 A flowchart of a product recommendation method based on a marketing model in one embodiment; Figure 3 Another schematic diagram of a process of a product recommendation method based on a marketing model in one embodiment; Figure 4 It is a schematic diagram of an oot KS based on an LCE algorithm model of a product recommendation method based on a marketing model in one embodiment; Figure 5 A schematic diagram of oot AUC based on an LCE algorithm model of a product recommendation method based on a marketing model in one embodiment; Figure 6 A schematic diagram of an OOTS KS based on an LR scoring card in a product recommendation method based on a marketing model in one embodiment; Figure 7 A schematic diagram of OOt AUC based on an LR score card for a product recommendation method based on a marketing model in one embodiment; Figure 8 A structural block diagram of a product recommendation device based on a marketing model in one embodiment; Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0026] It should be understood that in the description of the present application, unless the context clearly requires otherwise, words such as "include", "comprises", and the like throughout the specification should be interpreted as including rather than being exclusive or exhaustive; that is, as including but not limited to.

[0027] It should also be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, "plurality" means two or more.

[0028] It should be noted that the terms "S1", "S2", etc. are only used for the purpose of describing the steps, and do not specifically refer to the order or sequence, nor are they used to limit the present application. They are only for the convenience of describing the method of the present application, and cannot be understood as indicating the order of the steps. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present application.

[0029] In the current domestic and international environment, family wealth management is quietly changing, shifting from real estate to financial assets. With the end of the transition period of the new asset management regulations, bank wealth management is gradually entering the era of net value, corresponding to which is the optimization and upgrading of product structure and the healthy development of industry norms. Faced with a wealth of wealth management products, how to tap high-potential customers for marketing while ensuring safety, stability and profitability has become an urgent problem to be solved.

[0030] According to the background technology, the existing technology has prominent limitations, such as being sensitive to multicollinearity of independent variables in the model, difficulty in feature engineering to handle variables with larger dimensions, and insufficient model stability, resulting in poor ability to discover high-potential customers.

[0031] In order to solve the above technical problems, the present application provides a product recommendation method, device, equipment and storage medium based on a marketing model. By processing and screening data samples multiple times, it is possible to solve the problem of high-dimensional data processing. A marketing model is constructed based on the LCE algorithm, and the screened data is used to train and verify the constructed marketing model. The prediction and differentiation capabilities of the model and the stability of the model can be improved, so as to improve the accuracy of screening of target user groups and thereby improve the effectiveness of product marketing.

[0032] The product recommendation method based on the marketing model provided in this application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the data processing platform set on the server 104 through the network, wherein the terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0033] In one embodiment, Figure 2 As shown in the figure, a product recommendation method based on a marketing model is provided. Figure 1 The terminal in is used as an example to illustrate, including the following steps: S1: Based on a preset sample template, a first sample data set is obtained, and the first sample data set is preprocessed to obtain a second sample data set.

[0034] It should be noted that the sample template generally includes preset sample fields, data types of sample fields and data constraints, where the sample fields can be data such as name, age, occupation, etc., the data types of sample fields can be text, numbers, dates, etc., and the data constraints can be length restrictions, format requirements and numerical values, etc.; the sample data set can refer to customer data in the financial industry.

[0035] In some specific implementations, based on a preset sample template, obtaining a first sample data set includes: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, multiple target sample data are screened from the database, such as the sample range including: (1) aged 22-65 years old; (2) non-village bank customers; (3) non-bad customers; (4) completed first purchase; Classify the multiple target sample data to obtain positive sample data sets and negative sample data sets, wherein the target sample data are classified according to whether they are financial management customers. For example, new customers are defined as positive samples (labeled as 1), that is, customers who were not financial management customers last month but became financial management customers this month (average daily financial management amount is greater than 0), and customers who were not financial management customers last month but are not financial management customers this month are defined as negative samples (labeled as 1); Obtain the derivative data corresponding to the positive sample data set and the negative sample data set, and form a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data, wherein the asset data may include the customer's deposit assets, wealth management assets, AUM (Assets Under Management) and deposit transaction data, and the behavior data may include the data accumulated by the customer through the use of the APP, such as the number of logins in the past 30 days, the number of failed deductions in the past 90 days, the most recent login time, etc. The third-party data may include multiple data provided by various institutions, which is mainly used to measure the customer's willingness to repay, such as the number of loan applications in the past 6 months, the maximum credit limit in the past year, the maximum number of overdue days in the past 6 months, etc. The due diligence data is the data filled in by the user, such as the industry to which the work belongs, income and other information; Based on the positive sample data set and the negative sample data set and their corresponding derivative data, the first sample data set is generated, wherein the first sample data set may be a long list of about 1,700 variables generated based on the above data acquisition conditions.

[0036] In some specific implementations, preprocessing the first sample data set to obtain the second sample data set includes: The derived data in the first sample data set is cleaned to generate the second sample data set, wherein data cleaning generally involves checking the consistency of the data, removing dirty data such as missing values, single values, and duplicate values, and screening some invalid features such as ID, etc. Here, data variables with a high missing rate are removed.

[0037] In some specific implementations, after obtaining the second sample data set, the method further includes: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin, wherein for continuous variables, binning is usually required to convert them into categorical variables. In this application, missing values ​​and outlier data are binned separately, and other data are binned based on a binning method. The binning method generally includes equal frequency binning, equal interval binning, binning based on a decision tree or clustering algorithm, etc. The specific binning method can be selected according to actual needs; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical features of each bin are calculated and determined, that is, the character data is converted into numerical data, so as to facilitate modeling. The calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity, that is, whether they are monotonically increasing or monotonically decreasing within the defined interval. The bins corresponding to missing values ​​do not need to consider whether they satisfy monotonicity. In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, wherein the information value is used to measure the predictive ability of the feature, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin; After calculating the target information value, the method further includes: screening out data corresponding to when the target information value is greater than or equal to a second preset threshold, that is, eliminating data with poor effect, wherein the second preset value can be determined according to the content of Table 1, and can generally be 0.1.

[0038] Table 1: Mapping table of the range of information value and its corresponding prediction effect.

[0039] Range of information value Prediction effect Less than 0.02 Almost no 0.02-0.1 weak 0.1-0.3 medium 0.3-0.5 powerful Greater than 0.5 Need to confirm In the above implementation, when processing data with a large proportion of missing values, the data can be effectively utilized and the missing part can be treated as a bin. When processing outliers, the outlier data can be binned to further improve the robustness of the features. In addition, nonlinear features can be converted into linear features and variables of different dimensions can be aligned to the same scale, thereby improving the efficiency of model training.

[0040] S2: Obtain stability index values ​​of data in the second sample data set, and retain data corresponding to when the stability index value of the data is less than or equal to a first preset threshold, so as to generate a third sample data set.

[0041] It should be noted that the stability index value can be used to judge the stability of data features.

[0042] In some specific implementations, obtaining the stability index value of the data in the second sample data set, retaining the data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate the third sample data set includes: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, that is, the data screened out through the above steps, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, data corresponding to the stability index value is retained, wherein the first preset value can be determined according to the content of Table 2, and can generally be 0.1.

[0043] Table 2: Range of stability index values ​​and their corresponding stability mapping table.

[0044] Range of stability index values stability 0-0.1 good 0.1-0.25 Slightly unstable Greater than 0.25 Unstable The data set of the selected time period is further divided, such as selecting data from three time periods: 202108-202109, 202111-202112 and 202202-202203, the training set observation period 202108-prediction period 202109, observation period 202111-prediction period 202112, and the validation set observation period 202202-prediction period 202203. Because the amount of data at the observation points is large and cannot be used all for training, positive and negative samples can be sampled at a ratio of 1:10, such as 5200 positive samples and 52000 negative samples, that is, the retained data is sampled according to the preset ratio to generate a third sample data set, where the preset ratio can be set according to actual needs. The number of features retained based on the above screening is generally less than 50. For example, there are 4 types of data, and these 4 types of data have a total of 2000 variables. After processing through the above steps, the variables with the best prediction effect will be selected, such as 31.

[0045] In the above implementation, by determining the sample definition, processing the data, selecting appropriate observation points, completing data organization and cleaning, etc., from a long list to a short list for modeling, dividing the training set and the validation set, in order to screen out highly robust data from high-dimensional data, thereby improving the effect of model training, and further improving the model's predictive and distinguishing capabilities.

[0046] S3: constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard.

[0047] It should be noted that the generalization ability of an algorithm is often measured by bias-variance. Usually, the desired goal when training a model is to minimize bias and variance. Bagging (Bootstrap aggregating) divides a data set into multiple data sets by sampling with replacement, and trains multiple models separately, which plays a major role in reducing variance. In classification problems, voting is performed according to the principle of majority rule. In regression problems, the average of multiple test results is calculated. Boosting (a machine learning algorithm that can be used to reduce bias in supervised learning) uses the same weak learners as Bagging, but each new learner depends on the previous model and is combined according to a certain strategy, which plays a major role in reducing bias. The new integration method Local Cascade Ensemble (LCE) combines bagging-boosting methods to deal with the variance-bias balance in the model. Its characteristics are as follows: it uses a set of predictors in sequence and adds new attributes to the input data set at each stage; it adopts a divide-and-conquer strategy to apply cascade generalization locally, and reduces the bias of decision trees by using base-boosted predictors as basic learning. In the process of generating trees, the output of the base learner at each node is added to the data set as a new attribute to propagate the boost down the tree. The predicted output indicates the ability of the base learner to correctly predict the sample. At the next tree level, the output added to the data set is weighted by the base learner, so that more attention can be paid to samples that were previously incorrectly predicted. Finally, bagging is used to alleviate the overfitting caused by the boosted tree. Bagging reduces variance by creating multiple predictors from sampling and replacing the original data set to aggregate the tree with a simple majority vote.

[0048] In some specific implementations, constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, in response to detecting that the verification result of the first marketing model meets a preset standard, outputting a second marketing model includes: Based on the LCE algorithm, an initial marketing model is constructed, and the parameters of the initial marketing model are adjusted to obtain the first marketing model, wherein the parameters of the initial marketing model generally include: (1) base_estimator: basic classifier, the default is decision tree, you can choose different basic classifiers according to task requirements, such as support vector machine, random forest, etc.; (2) n_estimators: the number of models, that is, the number of cascaded model layers, which controls the complexity and performance improvement of the cascaded model. Usually, setting more base classifiers can improve the performance of the model, but it will also increase the computational overhead and may cause overfitting. Therefore, the setting value of this number is shown in Table 3; (3) K: Set the number of K nearest neighbors for building a neighborhood graph. This parameter affects the size of the neighborhood of local data points and can affect the generalization ability of the model. Therefore, this value is set to 10. (4) max_layers: sets the maximum number of cascade layers to avoid overfitting or performance degradation caused by excessive increase in the number of layers. It limits the maximum number of layers of the cascade model to avoid the model being too complex. The setting value of this number is shown in Table 3. (5) min_samples_neighbors: The minimum number of samples used for training each basic classifier in each cascade layer. It can control the scale of training samples in each cascade layer to avoid overfitting in the case of few samples. It determines the minimum number of sample data required for learning at each model layer to ensure that each group has enough resources to train the model. The setting value of this number is shown in Table 3. (6) verbose: controls the output detail level. You can set different values ​​to control the amount of information output during training, ranging from silent to detailed, to help monitor the progress of model training. When verbose=0, usually no information about the training process is output, and only the final result is returned after the training is completed. When verbose=1, usually a training progress bar for each epoch (period / round) and some basic information such as the loss value and accuracy rate on the training set are output.

[0049] Table 3: Parameter value table.

[0050] parameter explain value n_estimators The number of random forest estimators in each cascade 100 max_layers The number of favorite layers allowed by the cascader 6 min_samples_neighbors Minimum number of samples used to determine local neighborhoods 12 Based on the training set and the validation set in the third sample data set, the first marketing model is trained and validated respectively, wherein the third sample data set can be divided according to a preset ratio to generate the training set and the validation set, and the preset ratio can be 8:2, 7:3, etc.; In response to detecting that the verification result meets the preset standard, the second marketing model is output, which is the final marketing model, wherein the preset standard may be an accuracy rate higher than a certain value or a training number reaching a value, etc., and the method for determining the value can be set according to actual needs.

[0051] In some specific implementations, in order to verify the effect of the model constructed based on the LCE algorithm, after obtaining the second marketing model, the method further includes: The first evaluation index value and the second evaluation index value of the second marketing model are obtained, wherein the first evaluation index value and the second evaluation index value are index values ​​obtained using AUC (Area Under Curve, which is the area under the ROC curve and the coordinate axis) and KS (Kolmogorov-Smirnov, an important statistic for measuring the distinguishing ability of the model), respectively, and the calculation formulas are as follows: ; ; ; ; in, 、 、 、 The value of can be obtained from Table 4, represents the number of positive samples, represents the number of negative samples, represents the position (rank) of the i-th positive sample after sorting, represents the true positive rate, Represents the false positive rate.

[0052] Table 4: Comparative analysis table.

[0053] Predicted as positive class Predicted as negative class Actually positive TP FN Actually negative FP TN Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated. Before the evaluation, the LR score card needs to be used at the same time to perform model training and evaluation on the same sample to obtain the corresponding prediction results. The test results obtained by this application and based on the LR score card are as follows: Figure 4-Figure 7The results are shown in Table 5, where oot represents the trend result, % of population represents the threshold of the model prediction probability, and % of totalGood / Bad represents the ratio of the cumulative number of good customers to the total number of good customers (good%) and the cumulative number of bad customers to the total number of bad customers (bad%) in each scoring interval. The experimental results show that compared with the LR scoring card, the marketing model based on the LCE algorithm has improvements in KS and AUC, which makes the model have better predictive and distinguishing capabilities.

[0054] Table 5: Comparison of experimental results.

[0055] Model KS (oot) AUC(oot) LCE algorithm model 0.9284 0.6697 LR score card 0.8835 0.6253 S4: Acquire real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

[0056] It should be noted that real-time customer information can be obtained through the customer relationship management system, which may include its sample data and its derivative data. After processing the data in the same way as above, it can be input into the second marketing model. It can then be determined based on the output results whether the target user is a high-intention customer. If the target user is a high-intention customer, the corresponding marketing scenario will be configured for him to recommend product information. At the same time, the marketing effect can be detected based on the report for reference by the staff.

[0057] In the above-mentioned product recommendation method based on the marketing model, the method includes: based on a preset sample template, obtaining a first sample data set, and preprocessing the first sample data set to obtain a second sample data set; obtaining the stability index value of the data in the second sample data set, retaining the data corresponding to the stability index value of the data is less than or equal to the first preset threshold value to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, and outputting the second marketing model in response to detecting that the verification result of the first marketing model meets the preset standard; obtaining real-time customer information and inputting it into the second marketing model to determine the target user group, and recommending product information to the target user group. This application can solve the problem of high-dimensional data processing by processing and screening data samples multiple times, constructing a marketing model based on the LCE algorithm, determining the parameter selection of the LCE algorithm in the marketing scenario, and using the screened data to train and verify the constructed marketing model, which can improve the prediction and differentiation capabilities of the model and the stability of the model, so as to improve the accuracy of screening the target user group, thereby improving the effectiveness of product marketing.

[0058] It should be understood that although Figure 2-Figure 3The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-Figure 3 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0059] In one embodiment, Figure 8 As shown, a product recommendation device based on a marketing model is provided, comprising: a first data processing module, a second data processing module, a marketing model generation module and a user group determination module, wherein: A first data processing module, used to obtain a first sample data set based on a preset sample template, and preprocess the first sample data set to obtain a second sample data set; A second data processing module is used to obtain the stability index value of the data in the second sample data set, and retain the data corresponding to when the stability index value of the data is less than or equal to the first preset threshold value, so as to generate a third sample data set; a marketing model generation module, configured to construct a first marketing model based on the LCE algorithm, and train and verify the first marketing model using the third sample data set, and output a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; The user group determination module is used to obtain real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

[0060] As a preferred implementation, in the embodiment of the present invention, the first data processing module is specifically used for: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0061] As a preferred implementation manner, in the embodiment of the present invention, the first data processing module is further specifically used for: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0062] As a preferred implementation, in the embodiment of the present invention, the device further includes a screening module, and the screening module is specifically used for: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity; In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin.

[0063] As a preferred implementation, in the embodiment of the present invention, the second data processing module is specifically used for: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, retaining data corresponding to the stability index value; The retained data is sampled according to a preset ratio to generate a third sample data set.

[0064] As a preferred implementation, in the embodiment of the present invention, the marketing model generation module is specifically used for: Based on the LCE algorithm, an initial marketing model is constructed, and parameters of the initial marketing model are adjusted to obtain the first marketing model; Based on the training set and the validation set in the third sample data set, respectively training and validating the first marketing model; In response to detecting that the verification result meets the preset standard, a second marketing model is output.

[0065] As a preferred implementation, in an embodiment of the present invention, the device includes a model evaluation module, and the model evaluation module is specifically used to: Obtaining a first evaluation index value and a second evaluation index value of the second marketing model; Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated.

[0066] For the specific definition of the product recommendation device based on the marketing model, please refer to the definition of the product recommendation method based on the marketing model above, which will not be repeated here. Each module in the above-mentioned product recommendation device based on the marketing model can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be 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.

[0067] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device 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 non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a product recommendation method based on a marketing model is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0068] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0069] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: S1: Based on a preset sample template, a first sample data set is obtained, and the first sample data set is preprocessed to obtain a second sample data set; S2: Obtain stability index values ​​of data in the second sample data set, and retain data corresponding to when the stability index value of the data is less than or equal to a first preset threshold, so as to generate a third sample data set; S3: constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, and in response to detecting that the verification result of the first marketing model meets a preset standard, outputting a second marketing model; S4: Acquire real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

[0070] In one embodiment, when the processor executes the computer program, the processor further implements the following steps: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0071] In one embodiment, when the processor executes the computer program, the following steps are also implemented: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0072] In one embodiment, when the processor executes the computer program, the following steps are also implemented: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity; In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin.

[0073] In one embodiment, when the processor executes the computer program, the processor further implements the following steps: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, retaining data corresponding to the stability index value; The retained data is sampled according to a preset ratio to generate a third sample data set.

[0074] In one embodiment, when the processor executes the computer program, the processor further implements the following steps: Based on the LCE algorithm, an initial marketing model is constructed, and parameters of the initial marketing model are adjusted to obtain the first marketing model; Based on the training set and the validation set in the third sample data set, respectively training and validating the first marketing model; In response to detecting that the verification result meets the preset standard, a second marketing model is output.

[0075] In one embodiment, when the processor executes the computer program, the processor further implements the following steps: Obtaining a first evaluation index value and a second evaluation index value of the second marketing model; Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated.

[0076] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: S1: Based on a preset sample template, a first sample data set is obtained, and the first sample data set is preprocessed to obtain a second sample data set; S2: Obtain stability index values ​​of data in the second sample data set, and retain data corresponding to when the stability index value of the data is less than or equal to a first preset threshold, so as to generate a third sample data set; S3: constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, and in response to detecting that the verification result of the first marketing model meets a preset standard, outputting a second marketing model; S4: Acquire real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

[0077] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

[0078] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: The derived data in the first sample data set is cleaned to generate the second sample data set.

[0079] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity; In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin.

[0080] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, retaining data corresponding to the stability index value; The retained data is sampled according to a preset ratio to generate a third sample data set.

[0081] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: Based on the LCE algorithm, an initial marketing model is constructed, and parameters of the initial marketing model are adjusted to obtain the first marketing model; Based on the training set and the validation set in the third sample data set, respectively training and validating the first marketing model; In response to detecting that the verification result meets the preset standard, a second marketing model is output.

[0082] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: Obtaining a first evaluation index value and a second evaluation index value of the second marketing model; Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated.

[0083] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), 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.

[0084] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. A product recommendation method based on a marketing model, characterized in that: The method comprises: Based on a preset sample template, a first sample data set is acquired, and the first sample data set is preprocessed to obtain a second sample data set; Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to the first preset threshold, to generate a third sample data set; constructing a first marketing model based on the LCE algorithm, and training and verifying the first marketing model using the third sample data set, and outputting a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; Real-time customer information is obtained and input into the second marketing model to determine a target user group and recommend product information to the target user group.

2. The product recommendation method based on marketing model according to claim 1, characterized in that: Based on the preset sample template, obtaining the first sample data set includes: Build a sample template based on preset sample fields, data types of sample fields, and data constraints; Based on the sample template, a plurality of target sample data are screened out from a database; Classifying the multiple target sample data to obtain a positive sample data set and a negative sample data set; Obtaining derivative data corresponding to the positive sample data set and the negative sample data set, and forming a mapping relationship, wherein the derivative data includes at least one of the following: asset data, behavior data, third-party data, and due diligence data; The first sample data set is generated based on the positive sample data set and the negative sample data set and their corresponding derivative data.

3. The product recommendation method based on marketing model according to claim 2, characterized in that: Preprocessing the first sample data set to obtain a second sample data set includes: The derived data in the first sample data set is cleaned to generate the second sample data set.

4. The product recommendation method based on marketing model according to claim 1, characterized in that: Before obtaining the stability index value of the data in the second sample data set, the method further includes: Binning the derivative data in the second sample data set, and obtaining the number of positive sample data and the number of negative sample data in each bin; Based on the number of positive sample data and the number of negative sample data in each bin, the numerical feature of each bin is calculated and determined, and the calculation formula includes: ; in, represents numerical features, Indicates the proportion of negative sample data. Indicates the proportion of positive sample data; Check whether the numerical features of each bin satisfy monotonicity; In response to detecting that the numerical feature of each bin does not satisfy monotonicity, re-binning is performed and the numerical feature of each bin is calculated until the numerical feature of each bin satisfies monotonicity; In response to detecting that the numerical feature of each bin satisfies monotonicity, the information value of each bin is calculated, and the calculation formula includes: ; in, Indicates the information value of each bin; Based on the information value of each bin, the target information value is calculated and determined using the following formula: ; in, represents the target information value, Indicates the number of bins. Indicates The information value of each bin.

5. The product recommendation method based on marketing model according to claim 4, characterized in that: Obtaining stability index values ​​of data in the second sample data set, retaining data corresponding to when the stability index value of the data is less than or equal to a first preset threshold, to generate a third sample data set includes: The data corresponding to the target information value being greater than or equal to the second preset threshold value is screened out, and the stability index value is calculated, and the calculation formula is: ; in, represents the stability index value, Indicates that the training set is The actual proportion of each bin, Indicates that the validation set is The expected proportion of each bin; In response to detecting that the stability index value is less than or equal to a first preset threshold, retaining data corresponding to the stability index value; The retained data is sampled according to a preset ratio to generate a third sample data set.

6. The product recommendation method based on marketing model according to claim 1, characterized in that: Constructing a first marketing model based on the LCE algorithm, and using the third sample data set to train and verify the first marketing model, and in response to detecting that the verification result of the first marketing model meets the preset standard, outputting a second marketing model includes: Based on the LCE algorithm, an initial marketing model is constructed, and parameters of the initial marketing model are adjusted to obtain the first marketing model; Based on the training set and the validation set in the third sample data set, respectively training and validating the first marketing model; In response to detecting that the verification result meets the preset standard, a second marketing model is output.

7. The product recommendation method based on marketing model according to claim 1, characterized in that: After obtaining the second marketing model, the method further includes: Obtaining a first evaluation index value and a second evaluation index value of the second marketing model; Based on the first evaluation index value and the second evaluation index value, the effect of the second marketing model is evaluated.

8. A product recommendation device based on a marketing model, characterized in that: The device comprises: A first data processing module, used to obtain a first sample data set based on a preset sample template, and preprocess the first sample data set to obtain a second sample data set; A second data processing module is used to obtain the stability index value of the data in the second sample data set, and retain the data corresponding to when the stability index value of the data is less than or equal to the first preset threshold value, so as to generate a third sample data set; a marketing model generation module, configured to construct a first marketing model based on the LCE algorithm, and train and verify the first marketing model using the third sample data set, and output a second marketing model in response to detecting that the verification result of the first marketing model meets a preset standard; The user group determination module is used to obtain real-time customer information and input it into the second marketing model to determine the target user group and recommend product information to the target user group.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

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

Citation Information

Patent Citations

  • User screening method and device, computer equipment, storage medium and program product

    CN115545214A

  • Data processing method and system, storage medium and electronic equipment

    CN115879981A

  • Method for predicting short-term electricity price based on LCE algorithm

    CN117974216A

  • Method and system for assessing effectiveness of marketing campaigns using RFM matrix in real-time

    US20220414705A1