Model marketing suggestion generation method and device, electronic equipment and storage medium

Abstract

The invention discloses a model marketing suggestion generation method and device, electronic equipment and a storage medium in the technical field of computers. The method comprises the steps that marketing suggestion generation parameters are configured; checking whether the proportion of a positive sample to a negative sample in the marketing suggestion generation parameters meets a target proportion or not, and setting the marketing suggestion generation parameters meeting the target proportion as latest marketing suggestion generation parameters; obtaining latest marketing suggestion generation parameters, and generating marketing suggestions in one-to-one correspondence with the latest marketing suggestion generation parameters; and generating a marketing suggestion report according to the marketing suggestions. According to the method, a local interpretability algorithm is adopted, a refined marketing suggestion generation method and a marketable marketing suggestion system are provided, meanwhile, the contribution degree of customer features to a model result serves as the basis of marketing suggestions, marketing personnel can understand underlying logic for making marketing selection, and the marketing suggestion efficiency is improved. And the fit degree of the marketing suggestions and the actual conditions of the customers is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, electronic device and storage medium for generating model marketing suggestions. Background Art

[0002] Traditional bank marketing usually divides customer lists based on a single rule or a combination of rules, and only generates customer lists. Customer managers can only identify them based on their own experience combined with the lists, resulting in low efficiency.

[0003] In recent years, benefiting from the wide application of machine learning models in the field of marketing, the accuracy of customer marketing has been greatly improved. Currently, the technical solutions for common marketing scenarios in banks often adopt the method of formulating a single rule or a set of rules to screen marketing customer groups. The rules usually have two sources. One is to directly formulate screening rules based on business understanding; the other is to use simple interpretable models, such as machine learning models like logistic regression and decision trees. By establishing the relationship between customer characteristics and marketing business goals, the most important characteristics are selected for further analysis, and finally a series of rules are formed. Although rule-based marketing analysis is simple and has strong business interpretability, the depth of customer mining is often insufficient, and it is difficult to accurately match the most real needs of customers. At the same time, the rules of existing marketing programs are obtained from business experience, and the quality of marketing effects also completely depends on business experience. The reasons given by a single rule or a rule group mainly rely on the interpretability of the rule model itself. The explanation of the reasons for customer marketing is only limited to whether the rules are met, and it is impossible to further analyze the characteristics of customers to form clear model prediction reasons.

[0004] Even if a single rule or a set of rules can be regarded as a simple rule model with a certain degree of interpretability and can provide general reasons for decision-making, the general reasons provided are not formulated for the specific situations of each customer. Therefore, it is impossible to provide personalized information for specific customers, impossible to analyze the individual characteristics of customers, and lack of marketing pertinence for individual customers. On the other hand, complex machine learning models have weak interpretability and cannot give the reasons for making marketing choices, resulting in customer managers not understanding why they need to market the customers predicted by the model. There is often a lack of marketing focus in the marketing process, leading to a low marketing success rate. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, device, electronic device and storage medium for generating model marketing suggestions. By adopting a local interpretability algorithm, a refined method for generating marketing suggestions and an implementable marketing suggestion system are proposed. At the same time, taking the contribution degree of customer characteristics to the model result as the basis for marketing suggestions, it is possible to achieve refined analysis of the characteristics of each customer, which is beneficial for marketers to understand the underlying logic of making marketing choices and effectively improve the fit between marketing suggestions and the actual situation of customers themselves.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for generating model marketing suggestions, including configuring parameters for generating marketing suggestions;

[0008] Checking whether the ratio of positive samples to negative samples in the parameters for generating marketing suggestions meets the target ratio, and setting the parameters for generating marketing suggestions that meet the target ratio as the latest parameters for generating marketing suggestions;

[0009] Obtaining the latest parameters for generating marketing suggestions, and generating marketing suggestions corresponding one by one to the latest parameters for generating marketing suggestions;

[0010] Generating a marketing suggestion report according to the marketing suggestions.

[0011] As a further solution of the present invention: The method for configuring the parameters for generating marketing suggestions includes:

[0012] Obtaining the system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters input most recently, and setting the model probability threshold; taking the system parameters and interval threshold parameters as marketing target parameters; taking the customer feature partition parameters and model probability threshold parameters as customer feature parameters;

[0013] Dividing the customer feature partition parameters into continuous feature intervals or discrete feature intervals;

[0014] Dividing the feature samples included in the continuous feature interval into continuous positive samples or continuous negative samples, and dividing the feature samples included in the discrete feature interval into discrete positive samples or discrete negative samples;

[0015] Setting each code value in the discrete feature interval as a sub-discrete interval;

[0016] Using the method of equal-frequency or equal-distance binning to divide the value range of the continuous feature interval into several bins, and sequentially setting each bin as a sub-continuous interval;

[0017] Calculating the percentage of continuous positive samples in the feature samples in the continuous feature interval to obtain the positive sample rate;

[0018] Calculating the percentage of continuous negative samples in the feature samples in the continuous feature interval to obtain the negative sample rate;

[0019] Calculate the difference between the positive sample rate and the negative sample rate. When the difference is within the range of the target interval, the information entropy binning method is used to expand the difference between the positive sample rate and the negative sample rate; the difference between the maximum threshold and the minimum threshold in the target interval is close to 0. When the difference between the positive sample rate and the negative sample rate is within the target interval, it indicates that the values of consecutive positive samples and consecutive negative samples are close. The information entropy binning method is an interval partitioning method for continuous features in decision trees, with information entropy or information gain as the loss function. By adjusting the interval segmentation position, the information entropy of each interval after partitioning is minimized, and the sample purity of the samples obtained by the information entropy binning method is high;

[0020] Take the demarcation point between consecutive positive samples and consecutive negative samples as the output result of the marketing model; after inputting feature samples into the marketing model, when the positive sample rate generated by the marketing model is greater than or equal to the maximum threshold of the target interval, it indicates that the generation result of the marketing model is a positive sample. When the positive sample rate generated by the marketing model is less than the maximum threshold of the target interval, it indicates that the generation result of the marketing model is a negative sample;

[0021] Set the output result as the actual model probability threshold of the marketing model. By comparing the magnitude of the difference between the actual model probability threshold and the set model probability threshold, the difference degree between the customer feature parameters and the set model customer feature parameters can be obtained. When the difference between the actual model probability threshold and the set model probability threshold is large and exceeds the pre-set error threshold value, it is determined that the customer feature parameters do not conform to the current model customer features. By replacing the set model probability threshold and continuing to compare whether the difference between the set model probability threshold and the actual model probability threshold meets the threshold, the model customer features that the customer feature parameters conform to are determined.

[0022] As a further solution of the present invention: the method for obtaining the system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters of the most recent input includes:

[0023] Store the system parameters, interval threshold parameters, and model probability threshold parameters, and store the customer feature partition parameters in the way of feature name - interval number - position partition of the left and right endpoints of the interval. The number of customer feature partition parameters is 2 * number of features * number of intervals.

[0024] As a further solution of the present invention: the method for checking whether the ratio of positive samples to negative samples in the marketing recommendation generation parameters meets the target ratio includes:

[0025] Calculate the positive sample rate and negative sample rate in the customer feature partition parameters to obtain the first result;

[0026] Calculate the positive sample rate and negative sample rate in the model probability threshold parameters to obtain the second result;

[0027] Determine whether both the first result and the second result are significant;

[0028] If both the first result and the second result are significant, set the most recently input system parameter, interval threshold parameter, customer feature partition parameter, and model probability threshold parameter as the latest marketing recommendation generation parameters.

[0029] As a further solution of the present invention: The method for determining whether both the first result and the second result are significant includes:

[0030] Obtain the customer feature partition parameter of the initial business target and the model probability threshold parameter of the initial business target;

[0031] Obtain the customer feature partition parameter after the marketing target adjustment and the model probability threshold parameter after the marketing target adjustment;

[0032] Taking the model probability threshold of the initial marketing model as a benchmark, calculate the positive sample rate p of the customer feature partition parameter of the initial business target 0 ;

[0033] Taking the model probability threshold after the marketing target adjustment as a benchmark, calculate the positive sample rate p of the customer feature partition parameter after the marketing target adjustment;

[0034] When p ≥ p 0 , use the method of left - hand ratio hypothesis to test the first result; when p ≤ p 0 , use the method of right - hand ratio hypothesis to test the first result;

[0035] If the hypothesis is rejected, mark the marketing recommendations for the positive or negative samples of the customer feature partition parameter of the initial business target as not significant, and stop generating marketing recommendations;

[0036] If the hypothesis is accepted, mark the marketing recommendations for the positive or negative samples of the customer feature partition parameter of the initial business target as significant, and generate marketing recommendations.

[0037] As a further solution of the present invention: The method for generating marketing recommendations includes:

[0038] Input a sample recommendation description table and the most recently updated marketing recommendation generation parameters into a pre - trained marketing model; the marketing model can be any one of a neural network model, a decision tree model, or a logistic regression model;

[0039] Query the sample recommendation description table, find the corresponding recommendation description in the sample recommendation description table according to the positive or negative samples marked as significant, and obtain the group recommendation;

[0040] Query the sample suggestion description table, extract the feature values for the positive or negative samples marked as significant, find the corresponding quantiles on the sample suggestion description table according to the feature values, look up the suggestions corresponding to the quantiles in the sample suggestion description table, and generate individual suggestions;

[0041] Concatenate the group suggestions and individual suggestions to give marketing suggestions;

[0042] Drive manual review of marketing suggestions.

[0043] As a further solution of the present invention: The method for generating a marketing suggestion report according to the marketing suggestions includes:

[0044] Input the marketing model into the Shapley Additive Explanations (SHAP) algorithm program;

[0045] Input the customer feature partition parameters of the customer features into the marketing model. The marketing model determines whether the samples included in the customer feature partition parameters are positive samples or negative samples, and calculates the positive sample rate and negative sample rate;

[0046] When it is determined that the sample is a positive sample, calculate the contribution degree and select the K features with the largest contribution degree;

[0047] When it is determined that the sample is a negative sample, calculate the contribution degree and select the K features with the smallest contribution degree;

[0048] Successively obtain the customer feature partition parameters and quantiles corresponding to the K contribution degrees;

[0049] Obtain the group suggestions and individual suggestions corresponding to the quantiles, and use the combination of the group suggestions and individual suggestions as the marketing suggestions to send to the corresponding positions of the report template;

[0050] Judge whether the number of marketing suggestions is >= n;

[0051] When the number of marketing suggestions is less than n, judge as no, and then repeat the above steps of successively obtaining the customer feature partition parameters and quantiles corresponding to the K contribution degrees;

[0052] If the number of marketing suggestions is greater than or equal to n, judge as yes, and exit the loop to complete the generation of the report template;

[0053] Display the marketing suggestions and marketing suggestion reports of the specified customers on the front-end page, where n < K.

[0054] In a second aspect, there is provided a device that adopts the model marketing suggestion generation method as described in the above solution. The device includes;

[0055] A parameter configuration module, which is used to configure marketing recommendation generation parameters;

[0056] A parameter verification module, the input end of which is electrically connected to the output end of the parameter configuration module. The parameter verification module is used to verify whether the ratio of positive samples to negative samples in the marketing recommendation generation parameters meets the target ratio, and set the marketing recommendation generation parameters that meet the target ratio as the latest marketing recommendation generation parameters;

[0057] A marketing recommendation generation module, the input end of which is electrically connected to the output end of the parameter verification module. The marketing recommendation generation module is used to obtain the latest marketing recommendation generation parameters and generate marketing recommendations corresponding one by one to the latest marketing recommendation generation parameters;

[0058] A report generation module, the input end of which is connected to the output end of the marketing recommendation generation module. The report generation module is used to generate a marketing recommendation report based on the marketing recommendations.

[0059] In a third aspect, an electronic device is provided, including:

[0060] At least one processor; and a memory communicatively connected to the at least one processor;

[0061] Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the model marketing recommendation generation method as described in the above solution.

[0062] In a fourth aspect, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, it implements the model marketing recommendation generation method as described in the above solution.

[0063] Compared with the prior art, the beneficial effects of the present invention are:

[0064] 1. By adopting a local interpretability algorithm, the present invention proposes a refined marketing recommendation generation method and an implementable marketing recommendation system. At the same time, taking the contribution degree of customer characteristics to the model result as the basis of marketing recommendations, it can realize the refined analysis of the characteristics of each customer, and can identify and quantify the influence of each characteristic on the model prediction result, and then form targeted, data-driven marketing recommendations, providing deeper customer insight suggestions for marketers.

[0065] 2. The present invention can generate parameter dynamic adjustment suggestions based on updated marketing suggestions, can generate effective marketing suggestions and marketing suggestion reports by calculating whether the ratio of positive samples to negative samples meets the target ratio, which is conducive to enabling marketers to understand the underlying logic of making marketing choices, effectively improving the fit between marketing suggestions and the actual situation of customers themselves, facilitating marketers to grasp the key points of marketing, and improving the marketing success rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a flowchart of the method steps of the present invention;

[0067] Figure 2 is a flowchart of the method for generating a marketing suggestion report according to the marketing suggestion of the present invention;

[0068] Figure 3 is a block diagram of the device of the present invention;

[0069] Figure 4 is a structural diagram of an electronic device of the present invention;

[0070] Figure 5 is a display diagram of the suggestion report of the present invention.

[0071] In the figure: 1. Parameter configuration module; 2. Parameter inspection module; 3. Marketing suggestion generation module; 4. Report generation module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0072] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0073] Embodiment:

[0074] Please refer to Figure 1 , in an embodiment of the present invention, a method for generating a model marketing suggestion includes:

[0075] S1: Configure marketing suggestion generation parameters;

[0076] S2: Check whether the ratio of positive samples to negative samples in the marketing suggestion generation parameters meets the target ratio, and set the marketing suggestion generation parameters that meet the target ratio as the latest marketing suggestion generation parameters;

[0077] S3: Obtain the latest marketing suggestion generation parameters, and generate marketing suggestions corresponding one by one to the latest marketing suggestion generation parameters;

[0078] S4: Generate a marketing recommendation report based on the marketing recommendations.

[0079] Preferably, the method for configuring marketing recommendation generation parameters includes:

[0080] Obtain the system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters input most recently, and set the model probability threshold; use the system parameters and interval threshold parameters as marketing target parameters; use the customer feature partition parameters and model probability threshold parameters as customer feature parameters;

[0081] Divide the customer feature partition parameters into a continuous feature interval or a discrete feature interval;

[0082] Divide the feature samples included in the continuous feature interval into continuous positive samples or continuous negative samples, and divide the feature samples included in the discrete feature interval into discrete positive samples or discrete negative samples;

[0083] Set each code value in the discrete feature interval as a sub-discrete interval;

[0084] Adopt the method of equal-frequency or equal-distance binning to divide the value range of the continuous feature interval into several bins, and sequentially set each bin as a sub-continuous interval;

[0085] Calculate the percentage of continuous positive samples in the feature samples in the continuous feature interval to obtain the positive sample rate;

[0086] Calculate the percentage of continuous negative samples in the feature samples in the continuous feature interval to obtain the negative sample rate;

[0087] Calculate the difference between the positive sample rate and the negative sample rate. When the difference is within the interval range of the target interval, adopt the information entropy binning method to expand the difference between the positive sample rate and the negative sample rate; the difference between the maximum threshold and the minimum threshold in the target interval is close to 0. When the difference between the positive sample rate and the negative sample rate is within the target interval, it indicates that the values of the continuous positive samples and the continuous negative samples are close. The information entropy binning method is a method for the decision tree to divide the interval of continuous features, with the information entropy or information gain as the loss function. By adjusting the interval segmentation position, the information entropy of each divided interval is minimized, and the sample purity of the samples obtained by the information entropy binning method is high;

[0088] Take the demarcation point between the continuous positive samples and the continuous negative samples as the output result of the marketing model; after inputting the feature samples into the marketing model, when the positive sample rate generated by the marketing model is greater than or equal to the maximum threshold of the target interval, it indicates that the generation result of the marketing model is a positive sample. When the positive sample rate generated by the marketing model is less than the maximum threshold of the target interval, it indicates that the generation result of the marketing model is a negative sample;

[0089] Set the output result to the actual model probability threshold of the marketing model.

[0090] Preferably, the method for obtaining the system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters of the most recent input includes:

[0091] Store the system parameters, interval threshold parameters, and model probability threshold parameters, and store the customer feature partition parameters in the way of partitioning according to the position of feature name - interval number - left and right endpoints of the interval. The number of customer feature partition parameters is 2 * number of features * number of intervals.

[0092] Preferably, the method for checking whether the ratio of positive samples to negative samples in the marketing recommendation generation parameters meets the target ratio includes:

[0093] Calculate the positive sample rate and negative sample rate in the customer feature partition parameters to obtain a first result;

[0094] Calculate the positive sample rate and negative sample rate in the model probability threshold parameters to obtain a second result;

[0095] Judge whether both the first result and the second result are significant;

[0096] If both the first result and the second result are significant, set the system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters of the most recent input as the latest marketing recommendation generation parameters.

[0097] Preferably, the method for judging whether both the first result and the second result are significant includes:

[0098] Obtain the customer feature partition parameters of the initial business goal and the model probability threshold parameters of the initial business goal;

[0099] Obtain the customer feature partition parameters after the marketing goal adjustment and the model probability threshold parameters after the marketing goal adjustment;

[0100] Based on the model probability threshold of the initial marketing model, calculate the positive sample rate p of the customer feature partition parameters of the initial business goal 0 ;

[0101] Based on the model probability threshold after the marketing goal adjustment, calculate the positive sample rate p of the customer feature partition parameters after the marketing goal adjustment;

[0102] When p ≥ p 0 , use the method of left - hand ratio hypothesis to test the first result; when p ≤ p 0 , use the method of right - hand ratio hypothesis to test the first result;

[0103] If the hypothesis is rejected, the marketing suggestions for the positive or negative samples of the customer characteristic partition parameters of the initial business objective are marked as insignificant, and the generation of marketing suggestions is stopped;

[0104] If the hypothesis is accepted, the marketing suggestions for the positive or negative samples of the customer characteristic partition parameters of the initial business objective are marked as significant, and marketing suggestions are generated.

[0105] Preferably, the method for generating marketing suggestions includes:

[0106] Input the sample suggestion description table and the most recently updated marketing suggestion generation parameters into a pre-trained marketing model; the marketing model can be any one of a neural network model, a decision tree model, or a logistic regression model;

[0107] Query the sample suggestion description table, and according to the positive or negative samples marked as significant, find the corresponding suggestion description in the sample suggestion description table to obtain the group suggestion;

[0108] Query the sample suggestion description table, extract the feature values for the positive or negative samples marked as significant, find the corresponding quantiles on the sample suggestion description table according to the feature values, and find the suggestions corresponding to the quantiles in the sample suggestion description table to generate individual suggestions;

[0109] Concatenate the group suggestion and the individual suggestion to give the marketing suggestion;

[0110] Drive manual review of the marketing suggestions.

[0111] The marketing suggestions adopt a combination of group suggestions and individual suggestions. The feature interval corresponds to the group suggestion, and the feature values included in the feature interval correspond to the individual suggestions. The feature values obtain the corresponding quantiles according to the sample suggestion description table. For example:

[0112] When the customer characteristic partition parameter of a customer characteristic is the number of transactions in the past 7 days, and the value of the number of transactions is 90 times, then it is recommended to market the customer, which belongs to a positive sample; after dividing the customer characteristic partition parameter, this customer characteristic partition parameter falls into the interval of 80 - 100 times of transactions, and 90 times is the 75th percentile in the overall feature distribution; by querying the marketing suggestion table, the group suggestion corresponding to the interval of 80 - 100 times of transactions and the positive sample is "relatively high recent transaction frequency", and the individual suggestion corresponding to the 75th percentile is "exceeded 75% of customers", so the final generated suggestion is "relatively high recent transaction frequency, exceeded 75% of customers".

[0113] Preferably, as Figure 2 shown, the method for generating a marketing suggestion report based on the marketing suggestion includes:

[0114] Input the marketing model into the Shapley Additive Explanations (SHAP) algorithm program;

[0115] Input the customer feature partition parameters of the customer features into the marketing model. The marketing model determines whether the samples included in the customer feature partition parameters are positive samples or negative samples, and calculates the positive sample rate and the negative sample rate;

[0116] When it is determined that the sample is a positive sample, calculate the contribution degree and select the K features with the largest contribution degree;

[0117] When it is determined that the sample is a negative sample, calculate the contribution degree and select the K features with the smallest contribution degree;

[0118] Successively obtain the customer feature partition parameters and quantiles corresponding to the K contribution degrees;

[0119] Obtain the group suggestions and individual suggestions corresponding to the quantiles, and use the combination of the group suggestions and individual suggestions as the marketing suggestions and send them to the corresponding positions in the report template;

[0120] Judge whether the number of marketing suggestions is >= n;

[0121] When the number of marketing suggestions is less than n, judge it as no, and then repeat the above steps of successively obtaining the customer feature partition parameters and quantiles corresponding to the K contribution degrees;

[0122] If the number of marketing suggestions is greater than or equal to n, judge it as yes, and exit the loop to complete the generation of the report template;

[0123] Display the marketing suggestions and the marketing suggestion report of the specified customer on the front-end page, where n < K, as Figure 5 shown in the display of the marketing suggestions and the marketing suggestion report of the specified customer on the front-end page.

[0124] In a second aspect, as Figure 3 shown, a device is provided that uses the model marketing suggestion generation method of the above solution. The device includes;

[0125] A parameter configuration module 1, which is used to configure the marketing suggestion generation parameters;

[0126] A parameter verification module 2, the input end of the parameter verification module is electrically connected to the output end of the parameter configuration module. The parameter verification module is used to verify whether the ratio of positive samples to negative samples in the marketing suggestion generation parameters meets the target ratio, and set the marketing suggestion generation parameters that meet the target ratio as the latest marketing suggestion generation parameters;

[0127] A marketing suggestion generation module 3, the input end of the marketing suggestion generation module is electrically connected to the output end of the parameter verification module. The marketing suggestion generation module is used to obtain the latest marketing suggestion generation parameters and generate marketing suggestions corresponding one by one to the latest marketing suggestion generation parameters;

[0128] A report generation module 4, the input end of the report generation module is connected to the output end of the marketing suggestion generation module. The report generation module is used to generate a marketing suggestion report according to the marketing suggestions.

[0129] Thirdly, as Figure 4 shown, an electronic device is provided, including:

[0130] At least one processor; and, a memory communicatively connected to the at least one processor;

[0131] Wherein, the memory stores instructions executable by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can execute the model marketing suggestion generation method of the above solution.

[0132] Fourthly, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the model marketing suggestion generation method of the above solution is implemented.

[0133] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent replacements or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for generating model marketing suggestions, characterized in that: include: Configure marketing proposal generation parameters; Check whether the ratio of positive samples to negative samples in the marketing suggestion generation parameters meets the target ratio, and set the marketing suggestion generation parameters that meet the target ratio as the latest marketing suggestion generation parameters; Acquire the latest marketing suggestion generation parameters, and generate marketing suggestions corresponding to the latest marketing suggestion generation parameters; Generate marketing recommendations report based on marketing recommendations.

2. The method for generating model marketing suggestions according to claim 1, characterized in that: The method for configuring marketing suggestion generation parameters includes: Obtain the most recently input system parameters, interval threshold parameters, customer feature partition parameters, and model probability threshold parameters, and set the model probability threshold; Divide customer feature partition parameters into continuous feature intervals or discrete feature intervals; The feature samples contained in the continuous feature interval are divided into continuous positive samples or continuous negative samples, and the feature samples contained in the discrete feature interval are divided into discrete positive samples or discrete negative samples; Each code value in the discrete feature interval is set as a sub-discrete interval; The value range of the continuous feature interval is divided into several boxes by using the equal frequency or equal interval binning method, and each box is set as a sub-continuous interval in turn; Calculate the percentage of continuous positive samples in the feature samples in the continuous feature interval to get the positive sample rate; Calculate the percentage of continuous negative samples in the feature samples in the continuous feature interval to get the negative sample rate; Calculate the difference between the positive sample rate and the negative sample rate. When the difference is within the target interval, use the information entropy binning method to expand the difference between the positive sample rate and the negative sample rate. The cutoff point between continuous positive samples and continuous negative samples is used as the output result of the marketing model; Set the output result as the actual model probability threshold for the marketing model.

3. The method for generating model marketing suggestions according to claim 2, characterized in that: The method for obtaining the most recently input system parameters, interval threshold parameters, customer feature partition parameters and model probability threshold parameters comprises: Store system parameters, interval threshold parameters and model probability threshold parameters, as well as customer feature partition parameters in the manner of feature name - interval number - position partition of the left and right endpoints of the interval.

4. The method for generating model marketing suggestions according to claim 2, characterized in that: The method for checking whether the ratio of positive samples to negative samples in the marketing suggestion generation parameters meets the target ratio includes: Calculate the positive sample rate and the negative sample rate in the customer feature partition parameter to obtain a first result; Calculate the positive sample rate and the negative sample rate in the model probability threshold parameter to obtain a second result; Determine whether both the first result and the second result are significant; If both the first result and the second result are significant, the system parameters, interval threshold parameters, customer feature partition parameters and model probability threshold parameters input most recently are set as the latest marketing suggestion generation parameters.

5. The method for generating model marketing suggestions according to claim 4, characterized in that: Methods for determining whether both the first result and the second result are significant include: Obtain customer feature partition parameters of the initial business target and model probability threshold parameters of the initial business target; Obtain customer feature partition parameters after the marketing target is adjusted and model probability threshold parameters after the marketing target is adjusted; Based on the model probability threshold of the initial marketing model, the positive sample rate p0 of the customer feature partition parameter of the initial business target is calculated; Taking the model probability threshold after the marketing target is adjusted as a benchmark, calculate the positive sample rate p of the customer feature partition parameter after the marketing target is adjusted; When p≥p0, the first result is tested by the method of the left side hypothesis of the ratio; when p≤p0, the first result is tested by the method of the right side hypothesis of the ratio; If the hypothesis is rejected, the marketing suggestions for the positive or negative samples of the customer feature partition parameters of the initial business target are marked as insignificant, and the generation of marketing suggestions is stopped; If the hypothesis is accepted, the marketing suggestions for the positive or negative samples of the customer feature partition parameters of the initial business target are marked as significant, and a marketing suggestion is generated.

6. The method for generating model marketing suggestions according to claim 5, characterized in that: The method for generating marketing suggestions comprises: Inputting a sample suggestion description table and the most recently updated marketing suggestion generation parameters into the pre-trained marketing model; Query the sample suggestion description table, find the suggestion description corresponding to the sample suggestion description table according to the positive sample or negative sample marked as significant, and obtain the group suggestion; Query the sample suggestion description table, extract the feature value of the positive sample or negative sample marked as significant, find the corresponding quantile on the sample suggestion description table according to the feature value, find the sample suggestion description table to obtain the suggestion corresponding to the quantile, and generate individual suggestions; Combine group suggestions and individual suggestions to give marketing suggestions; Drive manual review of marketing recommendations.

7. The method for generating model marketing suggestions according to claim 6, characterized in that: The method for generating a marketing suggestion report according to the marketing suggestion comprises: The marketing model was input into the Shapley additive interpretation algorithm program; The customer feature partition parameters of the customer features are input into the marketing model, and the marketing model determines whether the samples included in the customer feature partition parameters are positive samples or negative samples, and calculates the positive sample rate and the negative sample rate; When a sample is judged to be a positive sample, the contribution is calculated and the K features with the largest contribution are selected; When a sample is judged to be a negative sample, the contribution is calculated and the K features with the smallest contribution are selected; Obtain customer feature partition parameters and quantiles corresponding to K contribution levels in sequence; Obtain group suggestions and individual suggestions corresponding to the quantiles, and send the combination of group suggestions and individual suggestions as marketing suggestions to the corresponding position of the report template; Determine whether the number of marketing suggestions is >= n; When the number of marketing suggestions is less than n, the judgment is no, and the above steps of obtaining the customer feature partition parameters and quantiles corresponding to K contribution degrees are repeated in sequence; If the number of marketing suggestions is greater than or equal to n, it is judged as yes, and the loop is exited to complete the generation of the report template; Display the marketing suggestions and marketing suggestion reports of the specified customers on the front-end page, where n <K。 8. A device, using the model marketing suggestion generation method according to any one of claims 1 to 7, characterized in that: The device comprises: A parameter configuration module, the parameter configuration module is used to configure marketing suggestion generation parameters; A parameter verification module, wherein the input end of the parameter verification module is electrically connected to the output end of the parameter configuration module, and the parameter verification module is used to verify whether the ratio of positive samples to negative samples in the marketing suggestion generation parameters meets the target ratio, and set the marketing suggestion generation parameters that meet the target ratio as the latest marketing suggestion generation parameters; A marketing suggestion generation module, wherein the input end of the marketing suggestion generation module is electrically connected to the output end of the parameter verification module, and the marketing suggestion generation module is used to obtain the latest marketing suggestion generation parameter and generate a marketing suggestion corresponding to the latest marketing suggestion generation parameter; A report generation module, wherein the input end of the report generation module is connected to the output end of the marketing suggestion generation module, and the report generation module is used to generate a marketing suggestion report according to the marketing suggestion.

9. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the model marketing suggestion generating method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the model marketing suggestion generating method according to any one of claims 1 to 7 is implemented.

Images