CN117649293A - Asset retention promotion method and system for bank-oriented issuing clients - Google Patents

Abstract

The invention belongs to the technical field of big data, and particularly relates to an asset retention and promotion method and system for a bank-oriented issuing customer. The method specifically comprises the steps of customer data processing, feature engineering, model training, model calculation, marketing list outputting and the like. The client processing program specifically comprises the steps of determining variables of a predicted target so as to determine positive samples and negative samples of subsequent model training, determining target guest groups of a substitution client, selecting screening indexes, completing data extraction, constructing a model wide table, processing and cleaning data and the like. The present invention also includes a system for performing the foregoing method, and a computer-readable storage medium having stored thereon a computer program for performing the foregoing method.

Description

Asset retention promotion method and system for bank-oriented issuing clients

Technical Field

The invention relates to the technical field of big data, in particular to an asset retention and promotion method and system for a bank issuing customer.

Background

With the development of the digitization trend of the internet industry, the traditional operation mode of banks has faced great challenges, and the digitization operation of banking industry is also a necessary way. The bank issuing customer is one of high-value business customer groups of banks, and how to excavate high-potential issuing customers, promote the issuing retention and the asset value of the high-potential issuing customers, and accordingly increase the contribution rate of the high-potential issuing customers to the bank value is one of the important concerns of the banks.

At present, banking industry has the condition that a large number of issuing clients have low asset retention rate and contribution rate, and meanwhile, the accuracy rate of identifying the demands of the issuing clients and the marketing inquiry is low, and the operation and maintenance effects on the issuing clients are not ideal, so that deep ploughing on the asset value of the issuing clients is lacking. Therefore, aiming at the bank-originated customers, a method capable of accurately analyzing and targeting marketing is found through big data and analysis technology, so that an effective tool is provided for the bank-originated customers to reserve and promote the assets, and the method is a problem to be solved in the field.

Disclosure of Invention

Aiming at the situation, the invention aims to provide an asset retention promotion method and system for a bank agent client. And constructing a prediction model for the asset promotion of the client by adopting a data modeling mode, and finally outputting a prediction result of the asset retention promotion by the model through training the model on data. And giving the prediction result to a personal business management department to formulate a marketing strategy of a corresponding generation client, helping banks to develop accurate marketing, and establishing a complete marketing maintenance system and a closed-loop flow.

In order to achieve the above object, the present invention provides an asset retention promotion method for a bank agent client, comprising the steps of:

step S1: customer data processing;

step S101, calling a data warehouse to acquire client data, and selecting a month and day average AUM lifting of a substitute client as a classification index of a model compared with a substitute wage of the substitute client which is larger than a certain threshold index;

step S102, determining a variable of a predicted target, and determining a target AUM value of asset promotion in the next month, so as to determine a positive sample and a negative sample in subsequent model training;

step S103, selecting an observation time point so as to determine a target guest group of the substitution client;

step S104: selecting a plurality of influencing factors from client information of a target client agent client group as screening indexes;

step S105: aiming at the target customer group of the generation client determined in the step S103, completing data extraction from the data source according to the screening index determined in the step S104, and constructing a model wide table;

step S106: processing and cleaning the extracted data in step S105;

step S2: feature engineering, namely generating new features or derivative variables aiming at the customer data processed in the step S1, and selecting the features by adopting a correlation coefficient method;

step S3: model training, namely dividing the data which finish the step S106 and the step S2 into a training set and a testing set, wherein the training set accounts for 80% of the total data, the testing set accounts for 20%, training the client data of the training set by using a machine learning algorithm, verifying model effects in the testing set, comparing training accuracy of different models, and selecting a champion model by adopting a classification model evaluation method;

step S4: after the champion model is determined in the step S3, an observation time point is determined again, the step S103 is executed, a new prediction target client set is established, the step S106 and the step S2 are further carried out, and the obtained target client group data are brought into the champion model determined in the step S3, so that a classification result of the prediction set is obtained;

step S5: and outputting a marketing list, and forming and outputting the marketing list aiming at the prediction set classification result obtained in the step S4.

As the above method, the definition is as follows: the issuing customer refers to a mechanism or a company customer which opens an account at a bank to transact a wage card for staff, the bank issues wages, prizes and the like to the staff wage card in full amount according to a wage detail list provided by a consignor, and the customer for collecting the wage card is the issuing customer. AUM (Asset Under Management) refers to the asset management scale of a customer, which is used by banks to measure the value of the customer. The higher the AUM of a customer, the higher the customer's contribution to the bank. The customer month and day average AUM refers to: the sum of the current month and month average deposit balance of the client and the current month and month average investment, wherein the current month and month average deposit balance of the client account is = (Σthe current month and month average investment balance of the client account)/the current month and day, the current month and month average investment balance is = (Σthe total client investment of the client) and the current month and day, and the current month and day average AUM of the clients in different months is not necessarily the same. Customer deposit promotion means that personal shipping wages are deposited, and other products (such as financial accounting, funds and the like) are used for receiving the personal shipping wages, so that the promotion of customer assets is finally realized.

As a preferred method for promoting the retention of assets for bank-oriented agents, step S102 is to use the ratio of the difference value of the current month, month and day average AUM minus the current month, month and day average AUM promotion of the last month and customer to the agent amount of the last month, if the ratio is greater than 0.5, the positive sample in the subsequent model training is used, and the prediction classification variable is 1; the remaining samples were taken as negative samples with a prediction classification variable of 0.

Preferably, the method for promoting asset retention for bank-oriented clients includes the influencing factors in step S104, including client basic information, channel information, asset distribution, credit index, consumption level, behavior information, income level and preference information.

As a further preferable aspect, the customer basic information in step S104 includes age, gender, academic, mobile banking subscription, weChat banking subscription information; the channel information comprises various offline and online current month transaction amounts and numbers; the asset distribution comprises month and day balance information of demand deposit, regular deposit, financial products, fund products, noble metals and the like; the credit index comprises consumption credit repayment and house credit repayment information; the consumption level comprises the consumption amount and the number of the consumption channels of the client in the current month; the behavior information comprises the login times of a client mobile phone bank and the browsing times of financial products; the income level comprises the amount of the generation of the customer, the number of the sales and the amount of the average month; the preference information includes individual channel in-out status of the customer.

Preferably, in the method for promoting asset retention for bank agent clients, the data processing and cleaning in step S106 is to clean, convert and integrate the original data, and process the missing value and the abnormal value.

Preferably, in the method for promoting asset retention for bank-based clients, the correlation coefficient method in step S2 calculates the correlation coefficient of each feature attribute with respect to the target value and the threshold 20, and then obtains the feature attribute with the largest 20 correlation coefficients.

Preferably, the machine learning algorithm in step S3 refers to decision trees, random forests, light tgbm, and the classification model evaluation method in step S3 adopts confusion matrix as model evaluation index.

As a further preferable method, the champion model in step S3 is a light tgbm model.

Preferably, in the method for promoting asset retention for a bank-oriented client, in step S5, the output marketing list is in the form of a data file, data is pushed to a marketing platform, and the marketing platform receives the data and then is used for subsequent outbound, so that closed-loop data transfer is realized.

As another aspect of the present invention, the present invention provides a system for performing the above-mentioned asset retention promotion method for a banking agent client, including:

the acquisition module is used for: the data calling step S101 is used for executing the data calling, and the data information of the substitute-transmitting client is obtained from the database by calling the data warehouse;

and a data processing module: the client data processing step is used for executing the client data processing steps from S102 to S106, and the data extraction, the data processing and the cleaning are completed aiming at the data obtained by the obtaining module;

and (3) a characteristic module: the feature engineering is used for executing the step S2, generating new features or derivative variables aiming at the client data which is processed by the data processing module, and selecting the features;

model training module: the method comprises the steps of (1) performing model training in the step (3) and determining an evaluation and champion model;

model calculation module: the model calculation in the step S4 is used for executing, so that a classification result of the prediction set is obtained;

and an output module: and the marketing list output module is used for executing the step S5, and outputting the marketing list according to the classification result obtained by the model calculation module.

As a third aspect of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which when executed by a processor implements any of the above-described asset retention promotion methods for banking agent customers.

Compared with the prior art, the method and the system have the following beneficial effects:

1. big data are fully utilized, customer data are integrated, the system flow is simplified, the data isolation is solved, and the system redundancy construction is avoided.

2. Through the machine learning algorithm, accurate marketing is realized, and a large amount of labor cost is saved.

3. The whole process realizes closed loop, and ensures the safety of data.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and other drawings can be obtained according to these framing drawings without the need of creative efforts for a person skilled in the art.

Fig. 1 is a flowchart of step S1 customer data processing in an asset retention promotion method for a banking agent customer according to an embodiment of the present invention.

Fig. 2 is a full flow chart of an asset retention promotion method for a banking agent client in accordance with an embodiment of the present invention.

Fig. 3 is a system schematic diagram of a method for performing asset retention promotion for a banking agent client according to an embodiment of the present invention.

Detailed Description

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. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The asset retention promotion method for the bank alternate delivery clients comprises the following steps:

step S1: customer data processing; the method specifically comprises the following steps:

step S101, calling a bank data warehouse to acquire client data, and selecting a date and month AUM lifting of a substitute client as a classification index of a model compared with a substitute wage of the substitute client which is larger than a certain threshold index;

AUM (Asset Under Management) refers to the asset management scale of a customer, which is used by banks to measure the value of the customer. The higher the AUM of a customer, the higher the customer's contribution to the bank. The proxy payroll refers to that an organization or a company client opens an account at a bank to transact a payroll card for staff thereof, and the bank distributes payroll, bonus and the like into the staff payroll card thereof in full amount on behalf of the consignor according to a payroll detail table provided by the consignor. The invention focuses on the generation payroll which is received by the individual clients. The purpose is to keep the personal proxy wages, and to receive the personal proxy wages by other products (such as financial accounting, fund, etc.), so as to realize the promotion of customer assets.

Step S102, determining a variable of a predicted target, and determining a target AUM value of asset promotion in the next month, so as to determine a positive sample and a negative sample in subsequent model training;

this means that we want the predictive model to be able to accurately predict how many customers in the group of forwarder customers of a potential asset promotion will be able to complete the asset promotion during the period of one month after the modeling time window. And subtracting the difference value of the current month, month and day AUM and the current month and day AUM and comparing the difference value of the current month, month and day AUM and the current day and day AUM and comparing the current generation amount ratio of the current month and day AUM to be more than 0.5, namely a positive sample in the subsequent model training. The remaining samples were taken as negative samples. Quantization index= (month and month date all AUM-month and month date all AUM last month and month date)/total amount of wage sent in last month. And if the final equivalence index is greater than 0.5, the sample is a prediction classification variable of 1, otherwise, the sample is 0.

Step S103, selecting an observation time point so as to determine a target guest group of the substitution client;

taking 28 days of 02 month of 2023 as a time division point of modeling, observing the generation clients of six months before the time point as a target client group, and predicting that the client of the next month promotes AUM value by more than half of the current generation wages.

Step S104: selecting a plurality of influencing factors from client information of a target client agent client group as screening indexes;

the potential influencing factors of the customer asset promotion are mainly analyzed from eight dimensions of the basic information, channel information, asset distribution, credit condition, consumption level, behavior information, income level and preference information of the customer. The basic information mainly related to clients comprises: age, gender, academy, mobile banking subscription, weChat banking subscription. The channel information includes: the amount and the number of transactions in the current month in various channels such as Jiang Duo, beijing Dong, payment device, mei Tuo and the like. The asset distribution includes: daily balance of living, regular, financial, fund, noble metal and the like. The credit index mainly comprises: and information such as payment of the consumption credit, payment of the house credit and the like. The consumption level comprises: the amount and number of the customers consume in each channel in the current month. The behavior information mainly comprises: the number of client mobile phone bank logins, the number of financial product browsing times and the like. The revenue level mainly comprises: the amount of the customer's generation, the number of the pen, the average amount of the month, etc. The preference information mainly includes: the various channels of the customer are switched in and out. Totally, 62 indexes are screened out. These indicators are all important factors affecting customer asset promotion.

Step S105: aiming at the target customer group of the generation client determined in the step S103, completing data extraction from the data source according to the screening index determined in the step S104, and constructing a model wide table;

after the influencing factors are selected, each dimension covers important information related to the assets of the foundry. And then, completing data extraction according to the dimensions, extracting data of each dimension of the issuing customer asset which is convenient to analyze from each data source, and unifying format caliber of the data. After the generation clients with the observation points of nearly six months are determined, searching for the logic relation among indexes in the data storage device, and extracting the data information of the clients in the bank by using a data processing tool. By data collection and extraction of these dimensions, we can build a comprehensive and detailed model-wide table containing all relevant data needed for the predictive model for the development of the customer's assets. This will provide a powerful support for subsequent modeling analysis and a valuable data base for providing a deeper understanding and predicting the potential of the forwarder.

Step S106: processing and cleaning the extracted data in step S105;

after the customer's portion of the data is obtained, the data is processed and cleaned. The method comprises the steps of cleaning, converting and integrating the original data, and mainly focuses on the problems of missing values, abnormal values, correlation and the like in the data. Statistical and visual analysis of the data is needed to understand the distribution, relevance and anomaly of the data and to handle missing values. And (3) filling a small amount of missing values in the data set by using a latest population method, and directly deleting the index for a large amount of missing value data.

Step S2: feature engineering, namely generating new features or derivative variables aiming at the customer data processed in the step S1, and selecting the features by adopting a correlation coefficient method;

the feature engineering is to generate new features or derivative variables through business experience and analysis of processing data, and the features are applied to model construction and improve training effect of the model. Feature selection is to select the most valuable feature for model construction to reduce dimensionality and improve model performance, and good data and features are the precondition for the model and algorithm to play a greater role. However, the existence of too many feature attributes may reduce the efficiency of model construction, and the effect of the model may be poor, so that the feature attribute with the greatest influence needs to be selected from the feature attributes as a feature attribute list of the last constructed model. Features that most impact the model output result and accuracy are identified by analyzing the correlation between variables, particularly with the target variables. Here we use a correlation coefficient method for feature selection. Correlation coefficient method: the correlation coefficient of each characteristic attribute to the target value and the threshold 20 are calculated, and then the characteristic attribute with the largest correlation coefficient of 20 is obtained. Therefore, the feature quantity is reduced, the modeling time is shortened, and the modeling efficiency is improved. The irrelevant features are removed, key factors are left, the difficulty of learning tasks is reduced, rules carried by data are more easily mined, and dimension disasters are reduced.

Step S3: model training, namely dividing the data which finish the step S106 and the step S2 into a training set and a testing set, wherein the training set accounts for 80% of the total data, the testing set accounts for 20%, training the client data of the training set by using a machine learning algorithm, verifying model effects in the testing set, comparing training accuracy of different models, and selecting a champion model by adopting a classification model evaluation method;

in the module, the most important feature set is screened out in combination with the earlier stage, and different model algorithms are applied to construct a classification model. In the modeling process, the data processed in the step S106 and the step S2 are divided into a training set and a testing set, wherein the training set accounts for 80% of the total data, and the testing set accounts for 20%. The customer data in the training set is then trained using a series of machine learning algorithm decision trees, random forests, LIGHTGBM (Light Gradient Boosting Machine). By training the model, a prediction result is obtained in the test set. The training accuracy of the different models was compared. The quality of a model is measured by corresponding evaluation indexes, but when different evaluation methods are adopted, corresponding evaluation results are different, a common classification model evaluation method is selected, and finally, a confusion matrix is preferably selected as the evaluation index of the model. Confusion matrices, which may also be referred to as likelihood tables or error matrices, are widely used in the machine learning field. It is a specific matrix, mainly used for evaluating whether model training performance is ideal or not. The rows of the matrix represent the actual categories, while the columns represent the predicted categories given by the model. This has the advantage that it is possible to intuitively reflect the result of whether or not there is confusion in the predicted class of the classification model, i.e. the classification that would have been attributed to this class into another class. TP is the number of samples for which the model prediction result is positive and the actual result is also positive; FP is the number of samples for which the model prediction result is positive but the actual result is negative; TN is the number of samples for which the model prediction result is negative and the actual result is also negative; FN is the number of samples for which the model prediction result is negative but the actual result is positive; tp+fp+fn+tn = total number of samples. Accuracy (Accuracy) indicates the proportion of the total samples occupied by the samples correctly predicted and classified by the classifier, and indicates the success rate of the model. And finally, selecting the LIGHTGGM model with higher accuracy as a champion model.

Step S4: after the champion model is determined in the step S3, an observation time point is determined again, the step S103 is executed, a new prediction target client set is established, the step S106 and the step S2 are further carried out, and the obtained target client group data are brought into the champion model determined in the step S3, so that a classification result of the prediction set is obtained; the method comprises the steps of carrying out a first treatment on the surface of the

And processing the predicted guest group by taking a predicted time point of which the 30 th year of 2023 is taken as a model, taking a client of which the time point is nearly half year as a target predicted guest group, and according to a mode-in characteristic established by a characteristic selection method in a training set and a data cleaning mode, so as to obtain index set data of the predicted set. Substituting the data into the training model to obtain the final classification result of the prediction set. At this time, the customer group with the classification result of 1 is the customer with the final desired potential improvement.

Step S5: and outputting a marketing list, and forming and outputting the marketing list aiming at the prediction set classification result obtained in the step S4.

And (3) adopting a result of the prediction model, carrying out accurate data insight on the existing clients according to crowd portrayal characteristics, and carrying out similarity product recommendation by combining purchase intention rules and past product preferences of the clients. After the data processing is finished, the data butt joint between the platforms is designed in order to realize the closed-loop online marketing of the client data. The high-potential value client data obtained by model prediction is located in a database, and the data is still in a background server at the moment, so that the marketing purpose is realized. If in the form of data export, the flow is cumbersome and costly. Therefore, the data is pushed to the marketing platform in the form of a data file, and after the accurate marketing platform receives the data, the accurate marketing platform can be used for subsequent outbound, so that closed-loop transmission of the data is realized. In the process, the safety of the data is ensured, and a great deal of labor cost is saved.

As an embodiment of the present invention, it is also apparent to provide a system for executing the above-mentioned asset retention promotion method for a banking proxy client, including: the acquisition module is used for: the data calling step S101 is used for executing the data calling, and the data information of the substitute-transmitting client is obtained from the database by calling the data warehouse; and a data processing module: the client data processing step is used for executing the client data processing steps from S102 to S106, and the data extraction, the data processing and the cleaning are completed aiming at the data obtained by the obtaining module; and (3) a characteristic module: the feature engineering is used for executing the step S2, generating new features or derivative variables aiming at the client data which is processed by the data processing module, and selecting the features; model training module: the method comprises the steps of (1) performing model training in the step (3) and determining an evaluation and champion model; model calculation module: the model calculation in the step S4 is used for executing, so that a classification result of the prediction set is obtained; and an output module: and the marketing list output module is used for executing the step S5, and outputting the marketing list according to the classification result obtained by the model calculation module.

In particular, according to the disclosed embodiments of the invention, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer software product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the methods shown in the flowcharts described above.

The embodiment of the disclosure also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the above-mentioned asset retention promotion method for a bank-oriented issuing customer.

By way of example, the readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The computer readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable storage medium may also be any readable medium that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.

Claims (10)

1. The asset retention promotion method for the bank alternate delivery clients is characterized by comprising the following steps:

step S1: customer data processing;

step S101, calling a data warehouse to acquire client data, and selecting a month and day average AUM lifting of a substitute client as a classification index of a model compared with a substitute wage of the substitute client which is larger than a certain threshold index;

step S102, determining a variable of a predicted target, and determining a target AUM value of asset promotion in the next month, so as to determine a positive sample and a negative sample in subsequent model training;

step S103, selecting an observation time point so as to determine a target guest group of the substitution client;

step S104: selecting a plurality of influencing factors from client information of a target client agent client group as screening indexes;

step S105: aiming at the target customer group of the generation client determined in the step S103, completing data extraction from the data source according to the screening index determined in the step S104, and constructing a model wide table;

step S106: processing and cleaning the extracted data in step S105;

step S2: feature engineering, namely generating new features or derivative variables aiming at the customer data processed in the step S1, and selecting the features by adopting a correlation coefficient method;

step S3: model training, namely dividing the data which finish the step S106 and the step S2 into a training set and a testing set, wherein the training set accounts for 80% of the total data, the testing set accounts for 20%, training the client data of the training set by using a machine learning algorithm, verifying model effects in the testing set, comparing training accuracy of different models, and selecting a champion model by adopting a classification model evaluation method;

step S4: after the champion model is determined in the step S3, an observation time point is determined again, the step S103 is executed, a new prediction target client set is established, the step S106 and the step S2 are further carried out, and the obtained target client group data are brought into the champion model determined in the step S3, so that a classification result of the prediction set is obtained;

step S5: and outputting a marketing list, and forming and outputting the marketing list aiming at the prediction set classification result obtained in the step S4.

2. The method for promoting the retention of assets for a bank-based on a client as claimed in claim 1, wherein the step S102 is characterized in that the variable for determining the prediction target is a ratio of a difference value obtained by subtracting the average AUM of the current month, month and day from the average AUM of the client of the previous month to the amount of the client of the previous month, and if the ratio is greater than 0.5, the positive sample in the subsequent model training is used, and the prediction classification variable is 1; the remaining samples were taken as negative samples with a prediction classification variable of 0.

3. The method for promoting asset retention for banking agent clients as claimed in claim 1, wherein the influencing factors in step S104 include client basic information, channel information, asset distribution, credit index, consumption level, behavior information, income level and preference information; step S104, the basic client information comprises age, gender, academic, mobile phone bank subscription and WeChat bank subscription information; the channel information comprises various offline and online current month transaction amounts and numbers; the asset distribution comprises month and day balance information of demand deposit, regular deposit, financial products, fund products, noble metals and the like; the credit index comprises consumption credit repayment and house credit repayment information; the consumption level comprises the consumption amount and the number of the consumption channels of the client in the current month; the behavior information comprises the login times of a client mobile phone bank and the browsing times of financial products; the income level comprises the amount of the generation of the client, the number of the strokes and the amount of the average month; the preference information includes individual channel in-out status of the customer.

4. The method for promoting asset retention for bank agent clients according to claim 1, wherein the data processing and cleaning in step S106 is to clean, convert and integrate the original data, and process the missing value and the abnormal value.

5. The method for promoting the retention of assets for a banking agent client according to claim 1, wherein the correlation coefficient method in step S2 calculates the correlation coefficient of each feature attribute with respect to the target value and the threshold 20, and then obtains the feature attribute with the largest 20 correlation coefficients.

6. The method for promoting asset retention for bank-based on client as claimed in claim 1, wherein the machine learning algorithm in step S3 is decision tree, random forest, light tgbm, and the classification model evaluation method in step S3 uses confusion matrix as model evaluation index.

7. The method for promoting asset retention for banking proxy clients as claimed in claim 6, wherein the champion model in step S3 is a light tgbm model.

8. The method for promoting the asset retention of the bank-oriented sending clients according to claim 1, wherein the step S5 of outputting the marketing list is to push data to the marketing platform in the form of a data file, and the marketing platform receives the data and then uses the data for subsequent outbound to realize closed-loop data transfer.

9. A system for performing the banking-agent-oriented asset retention promotion method of any one of claims 1-9, comprising:

the acquisition module is used for: the data calling step S101 is used for executing the data calling, and the data information of the substitute-transmitting client is obtained from the database by calling the data warehouse;

and a data processing module: the client data processing step is used for executing the client data processing steps from S102 to S106, and the data extraction, the data processing and the cleaning are completed aiming at the data obtained by the obtaining module;

and (3) a characteristic module: the feature engineering is used for executing the step S2, generating new features or derivative variables aiming at the client data which is processed by the data processing module, and selecting the features;

model training module: the method comprises the steps of (1) performing model training in the step (3) and determining an evaluation and champion model;

model calculation module: the model calculation in the step S4 is used for executing, so that a classification result of the prediction set is obtained;

and an output module: and the marketing list output module is used for executing the step S5, and outputting the marketing list according to the classification result obtained by the model calculation module.

10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the asset retention promotion method for a banking agent of any one of claims 1 to 8.