A bank financial product recommendation method, system, device and storage medium

Abstract

The present invention relates to the field of computer technology, and to a method, system, device and storage medium for recommending financial products for banks. The method comprises the following steps: collecting customer data and product data; dividing customer data into existing customer data and new customer data; making user portraits based on existing customer data and new customer data; making product portraits based on product data; establishing a first association relationship between the feature tags of existing customers and the product feature tags of products, and establishing a second association relationship between the feature tags of new customers and the product feature tags of products; outputting product recommendation results for existing customers based on the first association relationship, and outputting product recommendation results for new customers based on the second association relationship. The present invention can solve the problems that the existing bank product recommendation method cannot deeply mine customer data, relies on business scenarios and lacks interconnection and interaction between business scenarios, and the customer manager can only search and identify individually through the list, which is inefficient and low in accuracy.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, system, device and storage medium for recommending bank financial products. Background Art

[0002] With the rapid development of Internet technology, more and more Internet applications have taken root in various fields. The continuous innovation and development of Internet finance has intensified the fierce competition in the traditional financial market. Traditional financial institutions represented by commercial banks will face greater competitive pressure in financial marketing. To this end, commercial banks can learn from the technological means and marketing methods of Internet companies and apply them to financial marketing to cater to the current trend of Internet financial development. At present, commercial banks provide a wide variety of financial products and services, which inadvertently increases the cognitive burden of customers on products and services. The purpose of financial marketing is to provide products and services that meet customer conditions according to customer needs. In the context of inclusive finance, improving the level of precision marketing for different customer groups puts a test on the innovation ability of commercial banks themselves. Personalized recommendation system, as a recommendation technology that Internet companies have been practicing and receiving positive feedback from customers, has been widely recognized in the industry. Personalized recommendation system can make full use of customer data and mine customer behavior characteristics and interest preferences. It predicts products and services that customers may be interested in through the correlation between customer preferences and products, and then conducts precision marketing to customers.

[0003] At present, most banks use big data processing and rule analysis to generate a list of precision marketing customer groups. Some large banks have added complex scenario customer group labels such as machine learning models. Although the marketing customer groups formed based on traditional single rules or rule groups are simple to analyze and have strong business interpretability, they often cannot adapt to the changing business operation model and fail to effectively explore the potential relationships hidden in big data. First, traditional rule models cannot adapt well to the rapidly changing behavior patterns of customers. With the rapid development and application of technologies such as 5G, Internet of Things, big data, and cloud computing, new technologies and new scenarios emerge in an endless stream, and customers' behavior habits, preferences, and acceptance patterns are changing rapidly. At this time, rule mining is obviously too rough. Second, traditional rule models cannot fully extract valuable information from big data. At present, massive customer data is generated every day. Traditional rule models, even commonly used simple machine learning models such as logistic regression and decision trees, are usually difficult to capture complex nonlinear relationships and deep information in high-dimensional data when processing large-scale and high-dimensional data. Third, traditional rule models are more likely to miss valuable customers. Since traditional expert rules are more based on data statistics and business experience, the rule design is single and the indicator logic is simple, and the rules often cannot find all valuable customers. Fourth, the traditional rule model does not combine the laws of the customer life cycle with the customer value stratification, and cannot effectively conduct dynamic analysis of customers. Summary of the invention

[0004] The technical problem to be solved by the present invention is that the existing bank product recommendation method cannot deeply mine customer data, relies on business scenarios and lacks interconnection and interaction between business scenarios. Account managers can only search and identify individually through the list, which is inefficient and low in accuracy.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, a method for recommending banking financial products is provided, comprising the following steps: S1: collecting basic personal data of customers, customer transaction data and product data; S2: dividing the basic personal data of customers into basic personal data of existing customers and basic personal data of new customers; S3: dividing the customer transaction data into transaction data of existing customers and transaction data of new customers; S4: performing high-dimensional cross-precision positioning of the status of existing customers based on the basic personal data of existing customers, the customer group information of existing customers and the cycle value of existing customers, and performing customer profiling to obtain the customer portrait corresponding to each type of existing customer group; performing high-dimensional cross-precision positioning of the status of new customers based on the basic personal data of new customers, the customer group information of new customers and the cycle value of new customers, and performing customer profiling to obtain the customer portrait corresponding to each type of new customer group; and performing high-dimensional cross-precision positioning of the status of new customers based on the basic personal data of new customers, the customer group information of new customers and the cycle value of new customers, and performing customer profiling to obtain the customer portrait corresponding to each type of new customer group; and performing product value stratification based on product value stratification and product acceptance. S5: construct the existing customer feature labels according to the personal basic data of the existing customers, the customer group information of the existing customers, the cycle value of the existing customers and the customer portraits; construct the new customer feature labels according to the personal basic data of the new customers, the customer group information of the new customers, the cycle value of the new customers and the customer portraits; S6: establish a first association relationship between the existing customer feature labels and the product feature labels of the products, and establish a second association relationship between the new customer feature labels and the product feature labels of the products; S7: output the product recommendation results for the existing customers according to the first association relationship, including the existing customer product recommendation, the existing customer service recommendation and the existing customer message recommendation; output the product recommendation results for the new customers according to the second association relationship, including the new customer product recommendation, the new customer service recommendation and the new customer message recommendation.

[0007] Furthermore, after S1, the following steps are also included: preprocessing the collected customer personal basic data, customer transaction data and product data; the data preprocessing includes: cleaning invalid data and duplicate data; and standardizing the cleaned data.

[0008] Furthermore, after S3 and before S4, the following steps are also included: performing customer group attribution analysis based on the personal basic data of existing customers and the transaction data of existing customers to obtain the customer group information of existing customers; performing customer group attribution analysis based on the personal basic data of new customers and the transaction data of new customers to obtain the customer group information of new customers; performing life cycle positioning of existing customers based on the customer group attribution of existing customers, the personal basic data of existing customers and the transaction data of existing customers to obtain the life cycle of existing customers; performing life cycle positioning of new customers based on the customer group attribution of new customers, the personal basic data of new customers and the transaction data of new customers to obtain the life cycle of new customers; performing life cycle positioning of new customers based on the customer group attribution of new customers, the personal basic data of new customers and the transaction data of new customers to obtain the life cycle of new customers; performing life cycle positioning of new customers based on the customer group attribution of existing customers, the personal basic data of new customers and the transaction data of new customers to obtain the life cycle of new customers; performing life cycle positioning of new customers based on the customer group attribution of existing customers, the personal basic data of existing ... Based on the existing customer value stratification, the corresponding value stratification of existing customers is obtained; based on the new customer group affiliation, new customer personal basic data and new customer transaction data, the new customer value stratification is conducted to obtain the corresponding value stratification of new customers; based on the existing customer life cycle and the existing customer value stratification, the cycle value of existing customers is cross-analyzed to obtain the cycle value corresponding to each type of existing customer group; based on the new customer life cycle and the new customer value stratification, the cycle value of new customers is cross-analyzed to obtain the cycle value corresponding to each type of new customer group; based on the product data, the product value level is divided to obtain the product value stratification; based on the customer transaction data, the acceptance analysis is conducted to obtain the product acceptance.

[0009] Furthermore, customer life cycle positioning includes the following steps: dividing the customer life cycle into introduction, growth, maturity and decline; dividing customers into multiple customer groups according to customer personal basic data, customer transaction data and customer life cycle, including introduction customers, growth customers, maturity customers and decline customers; converting product recommendation models in different customer life cycles into each other.

[0010] Furthermore, customer value stratification includes the following steps: using the RFM model to stratify customer value; dividing customer value stratification into high value, medium value and low value; dividing customers into multiple customer groups, including high-value customers, medium-value customers and low-value customers, based on customer personal basic data, customer transaction data and customer value stratification.

[0011] Furthermore, the customer cross-analysis includes the following steps: modeling is performed using the obtained customer life cycle and customer value stratification to obtain a customer life cycle value curve.

[0012] Furthermore, customer profiling includes the following steps: using customer basic information, personal transaction data, customer group data and cycle value curve to conduct multi-dimensional analysis of customers to obtain customer status coordinates; using clustering algorithms to classify customer status coordinates; adding corresponding feature labels to each type of feature data.

[0013] Furthermore, product profiling includes the following steps: using customer transaction data to model product transaction status and obtain product acceptance; using product data to model product value and obtain product value stratification; using product data and product acceptance to classify through clustering algorithms; adding corresponding feature labels to each type of data.

[0014] Secondly, a banking financial product recommendation system is provided, including: a collection module for collecting customer personal basic data, customer transaction data and product data; a classification module for dividing customer personal basic data into existing customer personal basic data and new customer personal basic data, and dividing customer transaction data into existing customer transaction data and new customer transaction data; a product portrait module for dividing product value levels according to product data to obtain product value stratification; performing acceptance analysis according to customer transaction data to obtain product acceptance; performing product portraits according to product value stratification and product acceptance to obtain product feature labels corresponding to each type of product; an association module for The method comprises the following steps: constructing a feature label for existing customers based on the customer group information of new customers, the cycle value of existing customers and the customer portrait; constructing a feature label for new customers based on the personal basic data of new customers, the customer group information of new customers, the cycle value of new customers and the customer portrait; establishing a first association relationship between the feature label for existing customers and the product feature label for products, and establishing a second association relationship between the feature label for new customers and the product feature label for products; a recommendation module is used to output product recommendation results for existing customers based on the first association relationship, including product recommendations for existing customers, customer service recommendations for existing customers and message recommendations for existing customers; and outputting product recommendation results for new customers based on the second association relationship, including product recommendations for new customers, customer service recommendations for new customers and message recommendations for new customers.

[0015] Furthermore, the bank financial product recommendation system also includes: a data processing module for preprocessing the collected basic personal data of customers, customer transaction data and product data; a customer group positioning module for dividing existing customers into different customer groups to obtain the customer groups to which existing customers belong; for dividing new customers into different customer groups to obtain the customer groups to which new customers belong; a cycle value module for performing life cycle analysis and value stratification on existing customers to obtain the cycle value curve of existing customers; for performing life cycle analysis and value stratification on new customers to obtain the cycle value curve of new customers; a customer positioning portrait module The block is used to perform high-dimensional cross-precision positioning of the status of existing customers based on the personal basic data of existing customers, the customer group information of existing customers and the cycle value of existing customers, and to perform customer profiling to obtain the customer portrait corresponding to each type of existing customer group; the block is used to perform high-dimensional cross-precision positioning of the status of new customers based on the personal basic data of new customers, the customer group information of new customers and the cycle value of new customers, and to perform customer profiling to obtain the customer portrait corresponding to each type of new customer group; the data processing module includes: a data cleaning unit, used to clean invalid data and duplicate data; a standardization processing unit, used to perform standardization processing on the cleaned data.

[0016] Furthermore, the life cycle positioning module is used to divide the life cycle of existing customers according to the personal basic data of existing customers and the transaction data of existing customers to obtain the life cycle of existing customers; to divide the life cycle of new customers according to the personal basic data of new customers and the transaction data of new customers to obtain the life cycle of new customers; the life cycle positioning module includes: a period division module, used to divide the customer life cycle into an introduction period, a growth period, a mature period and a decline period; a customer group division module, used to divide customers into multiple customer groups, including introduction period customers, growth period customers, mature period customers and decline period customers, according to the collected personal basic data characteristics of customers, customer transaction data characteristics and customer life cycle characteristics; a value stratification positioning module, used to perform value stratification of existing customers according to the personal basic data of existing customers and the transaction data of existing customers to obtain the corresponding value stratification of existing customers; according to New customers' personal basic data and new customers' transaction data are used to perform value stratification on new customers to obtain corresponding value stratification for new customers; the value stratification positioning module includes: a value stratification module, which is used to divide customer value levels into high value, medium value and low value according to the RFM model; a customer group segmentation module, which is used to divide customers into multiple customer groups, including high-value customers, medium-value customers and low-value customers, based on the collected personal basic data characteristics, customer transaction data characteristics and customer value stratification characteristics; a life cycle curve drawing module, which is used to cross-analyze the life cycle value of existing customers based on the obtained life cycle of existing customers and the value stratification of existing customers to obtain the corresponding cycle value for each type of existing customer group; cross-analyze the life cycle value of new customers based on the new customer life cycle and the new customer value stratification to obtain the corresponding cycle value for each type of new customer group.

[0017] Furthermore, the coordinate construction unit performs high-dimensional cross-precision positioning of the status of existing customers based on the personal basic data of existing customers, the customer group information of existing customers and the cycle value to obtain the customer status coordinates; the data classification unit is used to classify the customer status coordinates using a clustering algorithm; the label adding unit is used to add corresponding feature labels to each type of feature data.

[0018] Furthermore, the product portrait module includes: an acceptance analysis unit, which is used to use customer transaction data to model product transaction status and obtain product acceptance; a value stratification unit, which is used to use product data to perform product value stratification and obtain product value stratification; a data classification unit, which is used to use product value stratification and product acceptance to perform classification through a clustering algorithm; and a label adding unit, which is used to add corresponding feature labels to each type of feature data.

[0019] According to a third aspect, a computer device is provided, comprising a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive data, and the processor is used to read the computer program to execute the bank financial product recommendation method as described in the first aspect.

[0020] In a fourth aspect, a computer-readable storage medium is provided, on which instructions are stored. When the instructions are executed on a computer, the bank financial product recommendation method as described in the first aspect is executed.

[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0022] 1. The algorithm can mine customers' behavioral habits, preferences, and acceptance patterns in big data. It can capture complex nonlinear relationships and deep information in high-dimensional data, and can better discover valuable customers and find more valuable customers.

[0023] 2. Relying on the composite design pattern of recommendation algorithm in customer life cycle management, different rule models at different stages of customer life cycle are incorporated into the overall framework of personal customer recommendation system. We can dynamically adjust the marketing focus according to the current business development stage.

[0024] 3. Based on the different business capabilities, task volumes and geographical environments of account managers, we provide a customer portrait function for secondary screening of the precision marketing list. By effectively adjusting the focus based on their actual situation and concentrating on breaking through valuable customers, marketing efficiency will be greatly improved.

[0025] 4. In combination with specific application scenarios, the rules and models of each stage of the customer life cycle are integrated into the overall framework of the recommendation algorithm to produce a phased precision marketing list; different rule models at different stages of the customer life cycle can be incorporated into the overall framework of the personal customer recommendation system, and the marketing focus can be dynamically adjusted according to the current business development stage.

[0026] 5. By relying on the dynamic positioning of customer status and dynamic product profiling, customer needs and products can be dynamically aligned to make more accurate recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 A schematic diagram of the execution flow of the bank financial product recommendation method provided in an embodiment of the present invention.

[0029] Figure 2 A schematic diagram of the execution principle of the bank financial product recommendation method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0031] In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is apparent to one of ordinary skill in the art that these specific details are not necessarily employed to practice the present invention. In other embodiments, in order to avoid obscuring the present invention, well-known structures, circuits, materials, or methods are not specifically described.

[0032] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment," "an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. In addition, it will be appreciated by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0033] In the description of the present invention, the directions or positional relationships indicated by terms such as “front”, “rear”, “left”, “right”, “up”, “down”, “vertical”, “horizontal”, “high”, “low”, “inside” and “outside” are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the scope of protection of the present invention.

[0034] Example:

[0035] The first aspect is to provide a banking and financial product recommendation method that integrates customer life cycle characteristics, customer value stratification characteristics, and banking and financial product characteristics, and innovatively designs a multi-stage recommendation algorithm based on a multi-model combination of customer life cycle and customer value stratification, effectively integrating multi-model capabilities to achieve more accurate and comprehensive marketing effects. On the basis of algorithm innovation, combined with the characteristics of precision marketing, a supporting customer portrait is designed, and a recommendation information query platform is built to ensure that marketing information is accurate, convenient, and comprehensive.

[0036] The customer life cycle theory evolved from the product life cycle theory. As an important resource of an enterprise, customers have important value and life cycle. The customer life cycle theory refers to the whole process from the establishment of a business relationship between an enterprise and a customer to the complete termination of the relationship. It is the development trajectory of the customer relationship level over time. It dynamically describes the overall characteristics of the customer relationship at different stages. The customer life cycle can be divided into four stages: introduction, growth, maturity and decline. Among them, the introduction stage is the gestation period of customer relationships, the growth stage is the rapid development stage of customer relationships, the maturity stage is the mature and ideal stage of customer relationships, and the decline stage is the stage when the customer relationship level reverses. As a very useful tool, the customer life cycle theory is an important method to gain insight into the dynamic characteristics of customer relationship development.

[0037] Customer value stratification is a method of distinguishing the value of customer groups and managing them in layers. It is based on the difference in the contribution of customers to the enterprise, and customers are classified and managed in layers according to the size of their value contribution, so as to help enterprises maximize customer value under limited resources. The customer value stratification theory believes that different customers have different contributions and needs to the enterprise. Customers can be divided into high-value customers, medium-value customers, and low-value customers based on indicators such as customer consumption behavior, loyalty, and profit contribution. Through this hierarchical management, enterprises can formulate differentiated service strategies for customers at different levels, focusing on maintaining the loyalty of high-value customers, improving the consumption potential of medium-value customers, and effectively managing the service costs of low-value customers. As an important tool for enterprise refined operations, customer value stratification helps enterprises achieve optimal resource allocation while improving customer satisfaction.

[0038] This embodiment first distinguishes the stage of the customer, sets different product customer access rules for customer groups at different stages, and then defines the recommendation target in combination with the marketing scenario. For example, for credit card products, the introduction period customers are defined as customers who have not activated, have not used their cards for the first time, and have a short account opening time. Rules are used to recommend the list of customers who have not activated the cards in the first three periods; the growth period customers are defined as customers whose transaction numbers are rising and who have not yet or rarely show key behaviors. Rules can be used to recommend a list of customers for specific products; mature period customers are defined as positive value customers who contribute more to the issuing bank. Association rules are used to predict the customer product demand list, which can predict changes in customer transaction behavior in advance and prevent customer dormancy and loss; decline period customers are defined as customers who have held their cards for a long time and have not used their cards recently, as well as customers who have applied for full account cancellation. Models can be used to predict the customer wake-up list. For these four stages of credit cards, combined with specific application scenarios, the rules and models of each stage are integrated into the overall framework of the recommendation algorithm to produce a staged precision marketing list.

[0039] By stratifying the value of customers, differentiated product recommendation strategies can be formulated for customers at different levels. First, high-value customers, medium-value customers, and low-value customers are divided according to the value contribution of customers, and then the recommendation strategy is defined in combination with actual business needs. Taking credit card products as an example, high-value customers refer to customers with frequent transactions, high consumption amounts, and strong loyalty. For such customers, high-limit credit cards or value-added services can be recommended, such as VIP channels, airport VIP lounges, and other exclusive benefits. Medium-value customers are those whose consumption is relatively stable, but whose potential value has not been fully explored. The recommendation strategy can increase the frequency and limit of their credit card use by pushing incentives such as installment payments and consumer cashback. For low-value customers, such customers spend less or have not used credit cards for a long time, their future potential behavior can be predicted through the model, and activation and awakening can be attempted by issuing coupons for use, free annual fees, and so on. In the application of credit card products, this stratified embodiment effectively improves the effect of precision marketing and optimizes customer resource allocation by integrating customer value stratification and recommendation models.

[0040] Different rule models and customer value stratification at different stages of the customer life cycle, after being incorporated into the overall framework of the personal customer recommendation system, can dynamically adjust the marketing focus according to the current business development stage. For example, if it is still in the incremental customer circle stage, more recommendation computing power should be applied to the introduction and growth stages of medium-value and low-value products. In addition, the recommendation capabilities between different stages can be transformed into each other. For example, cross-marketing recommendations for high-value users in the mature stage will enable customers to enter the introduction period of another financial product in a high-value state.

[0041] Based on the above technical ideas, the execution process of this method is as follows Figure 1 As shown, the following steps are included:

[0042] Step 1: Collect basic personal data of customers, including personal transaction data and other data.

[0043] Step 2: Analyze the customer segments based on their personal basic data.

[0044] Step 3: Calculate the customer’s life cycle value based on the customer’s personal basic data and customer group information.

[0045] Step 4: Create customer profiles based on their personal basic data, customer group information, and cycle value.

[0046] Step 5: Input the customer's basic personal data, customer group information, cycle value and customer portrait information into the recommendation engine and calculate the recommendation results, including recommended products, services or messages.

[0047] Based on the above technical ideas, the execution principle of this method is as follows Figure 2 As shown, the following steps are included:

[0048] Step 1: Collect customer basic personal data, customer transaction data and product data.

[0049] At a specified time, obtain a list of all target customers within the current banking system and obtain relevant data on the customer group, including basic information characteristics of the customers (such as age, gender, education, marital status, income, etc.), basic financial asset characteristics (such as deposit balance, loan balance, etc.), transaction behavior characteristics, etc.

[0050] Step 2: Preprocess the collected customer personal basic data, customer transaction data and product data.

[0051] Data preprocessing includes: cleaning invalid data and duplicate data; standardizing the cleaned data.

[0052] Banks have a large amount of customer data, but the data obtained directly from banks is difficult to use directly. The acquired data needs to undergo a series of processing, including cleaning invalid data, duplicate data, etc. The cleaned data also needs to be standardized.

[0053] Step 3: Divide customer personal basic data into existing customer personal basic data and new customer personal basic data, and divide customer transaction data into existing customer transaction data and new customer transaction data.

[0054] The basic personal data of existing customers and new customers can be divided according to the account opening time and the threshold of the number of products held. For example, for new customers, their basic information dimensions are fixed, and there is no derivative product and transaction data, so the actual number of products held is less than the threshold of the number of products held. Customer transaction data is divided according to the divided basic personal data of customers.

[0055] Step 4: Conduct customer group positioning based on the personal basic data and transaction data of existing customers to obtain the customer group to which the existing customers belong; conduct customer group positioning based on the personal basic data and transaction data of new customers to obtain the customer group to which the new customers belong.

[0056] Step 5: Perform a cycle value analysis on existing customers based on their personal basic data, transaction data, and customer group information to obtain the cycle value curve corresponding to existing customers; perform a cycle value analysis on new customers based on their personal basic data, transaction data, and customer group information to obtain the cycle value curve corresponding to new customers.

[0057] This step mainly includes customer life cycle segmentation and customer value stratification. Customer life cycle segmentation is to divide the entire customer life cycle into several stages by analyzing the behavioral characteristics and demand changes of customers at different stages, so as to achieve precise service and marketing. The life cycle of bank customers usually covers several important stages such as the customer's first contact, active period, mature period and churn period. By dividing the customer life cycle, banks can more clearly understand the financial needs of customers at different stages and the changes in their behavior patterns. For example, in the first contact stage of customers, customers' needs are often concentrated on basic financial products such as savings accounts and credit cards. As customers gradually enter the active period, they may begin to consider more diversified financial products such as loans or investment and financial management. When customers enter the mature period, banks can recommend high value-added products such as wealth management or customized financial services to them by analyzing their historical transaction data and behavioral habits. Finally, at the stage where customers may churn, banks need to take proactive measures, such as providing preferential or personalized services, to delay customer churn. By dividing the customer life cycle, not only can customer satisfaction be effectively improved, but it can also help banks optimize marketing strategies and achieve more targeted customer care and resource allocation.

[0058] Customer value stratification refers to the process of categorizing customers into different levels according to their value by analyzing their economic contribution, behavioral characteristics, and future potential, so that banks can implement targeted management and services. Since banks have limited resources, customer value stratification can help optimize resource allocation, improve marketing efficiency, and maximize customer lifetime value. When constructing customer value stratification, it is first necessary to determine which attributes can effectively measure the current value and future potential of customers. Common stratification criteria include customer account balances, monthly transaction amounts, product holdings, historical transaction behaviors, and customer dependence on bank products. For example, customers with high account balances and long-term bank wealth management products are usually classified as high-value customers. Customers who trade frequently but with small single transaction amounts may be classified as medium-value customers. At the same time, the stratification criteria can be further refined by analyzing customer activity, loyalty, risk preferences, etc. In order to achieve accurate customer value stratification, customers are usually divided into high, medium, and low-value customer groups by combining customer historical data and behavioral analysis. For high-value customers, banks can provide customized financial products and value-added services to maintain their loyalty; for medium-value customers, they can tap their potential value by increasing the frequency of interaction and personalized recommendations; and for low-value customers, banks can adopt standardized service models and increase their contribution through low-cost marketing methods. By stratifying customer value, banks can not only implement differentiated service strategies, but also increase the overall contribution rate of customers, ensuring effective maintenance and improvement of customer relationships in a highly competitive market.

[0059] Step 5.1: According to the life cycle theory, divide the customer life cycle into introduction, growth, maturity and decline stages.

[0060] Step 5.2: Based on the collected personal basic data characteristics, customer transaction data characteristics and customer life cycle characteristics, customers are divided into multiple customer groups, including introduction customers, growth customers, mature customers and decline customers.

[0061] First, we define data about customer lifecycle characteristics, including the time of the customer's first account opening, changes in customer account activity, and the timeline of product use. The time of first account opening refers to the time since the customer opened the account; changes in activity refer to the time series data based on account balance or transaction frequency, extracting the trend of customer activity over time; product usage time refers to the length of time a customer uses a specific bank product (such as a loan or credit card).

[0062] Based on customer life cycle characteristics, transaction behavior and personal foundation, K-means and other algorithms are used to cluster customers and obtain preliminary life cycle division results. Then the clustering results are modified using rule-based methods: introduction customers are defined as those with an account opening time of less than 3 months and a transaction frequency of <5 times / month; growth customers are defined as those with an account opening time of more than 3 months and less than 1 year, and a transaction frequency of 5-15 times / month; mature customers are defined as those with an account opening time of more than 1 year, a transaction frequency of >15 times / month, and the use of multiple products; decline customers are defined as those with a continuous decline in transaction frequency for 3 months or a long-term low account balance.

[0063] Step 5.3: Transform product recommendation models at different customer life cycles.

[0064] By regularly analyzing customer behavior data, we can identify when customers transition from one life cycle stage to another. When a customer enters a new life cycle stage, the conversion of the product recommendation model is triggered. We can make judgments based on rules and define the conversion conditions for each stage based on account activity, transaction frequency, product usage, etc. For example, if a customer's transaction frequency increases significantly for three consecutive months, it indicates that the customer may transition from the introduction stage to the growth stage. At the same time, we set a trigger mechanism for life cycle conversion. The system regularly scans all customers and calculates whether the characteristic values ​​of their life cycle stages meet the conversion conditions. Once the conversion conditions are triggered, the system converts the recommendation model that the customer currently belongs to to the recommendation model of the new life cycle stage.

[0065] Step 5.4: Use the RFM model to stratify customer value into high value, medium value and low value; divide customers into multiple customer groups, including high value customers, medium value customers and low value customers, based on the collected customer personal basic data characteristics, customer transaction data characteristics and customer value stratification characteristics.

[0066] The RFM model can quantitatively analyze customers through the three dimensions of Recency (last consumption time), Frequency (consumption frequency) and Monetary (consumption amount), and then judge the customer's value level. Recency (R) refers to the time interval between the customer's last transaction and the present, reflecting the customer's activity. The shorter the time interval, the more active the customer. Frequency (F) refers to the frequency of customer consumption in a specific time period. The higher the frequency, the more loyal the customer. Monetary (M): the amount of customer consumption in a specific time period. The higher the amount, the greater the customer's contribution to the company. R, F, and M are divided into five levels (such as from 1 to 5) according to the value size, and each level represents the relative ranking of the customer. For example, for the R score, 5 points means recent purchase; 1 point means no purchase for a long time; for the F score: 5 points means a large number of purchases; 1 point means a small number of purchases; for the M score: 5 points means a high total consumption amount; 1 point means a low total consumption amount. The scores of the three dimensions of R, F, and M are weighted and summed to obtain the comprehensive RFM score of each customer.

[0067] Based on the obtained R, F, and M, the K-means clustering algorithm is used to cluster customers, and the number of clusters is set to 3 (high, medium, and low). Based on the clustering results, the customer characteristics of each cluster are analyzed and marked as high-value, medium-value, and low-value customers, respectively, to obtain preliminary stratification results. Then, based on the RFM comprehensive score, the stratification results are fine-tuned using the rule method. Customers with a comprehensive RFM score of 80% or above can be adjusted to high-value customers, customers with a comprehensive RFM score between 30% and 80% can be adjusted to medium-value customers, and customers with a comprehensive RFM score of 30% or below can be adjusted to low-value customers.

[0068] Step 5.5: Use the obtained existing customer life cycle and existing customer value stratification to obtain the existing customer cycle value curve; use the obtained new customer life cycle and new customer value stratification to obtain the new customer cycle value curve.

[0069] Step 6: Based on the personal basic data of existing customers, transaction data of existing customers, customer group information of existing customers and the life cycle value curve of existing customers, perform high-dimensional cross-precision positioning of the status of existing customers, and conduct customer profiling to obtain the characteristic labels corresponding to each type of existing customer group; based on the personal basic data of new customers, transaction data of new customers, customer group information of new customers and the life cycle value curve of new customers, perform high-dimensional cross-precision positioning of the status of new customers, and conduct customer profiling to obtain the characteristic labels corresponding to each type of new customer group.

[0070] Customer portrait refers to the extraction and analysis of the characteristics of a single customer, and the construction of a portrait is achieved by labeling specific customers. Since bank products are targeted at a certain customer group rather than individual customers, grasping the characteristics of group customers can help clarify the needs of the customer group and help improve the efficiency of precision marketing. Customer group portraits are different from individual customer portraits, and their core focus is on macro groups. By obtaining customer characteristics from massive customer data, using data analysis to group customers with similar characteristics together, labeling the customer groups according to their characteristics, and forming group customer portraits. In terms of portrait design, this embodiment supplements historical marketing information, historical portrait indicator change information, recommended product lists and recommendation reasons, to help account managers fully obtain marketing data.

[0071] Step 6.1: Use clustering algorithm to classify customers.

[0072] By analyzing the cycle value curve, we can define the status of customers at different stages. For example, the introduction stage with high value indicates new customers with large transaction amounts and potential customers; the mature stage with medium value indicates long-term active customers with medium transactions but room for improvement; the decline stage with low value indicates customers with gradually decreasing interactions and low value contribution, who need to be awakened.

[0073] Using life cycle characteristics (such as activity, transaction frequency, etc.) and value stratification characteristics (such as RFM score), customers in different periods are clustered by K-means clustering method, and each group is marked by analyzing the clustering results. Then, the clustering results are corrected using the rule method, and the rules can be defined according to the life cycle and RFM score obtained in step 5. For example, the introduction period and high value can be defined as the account opening time is less than 3 months, R>4, F>4, M>4; the mature period and medium value can be defined as the account opening time is greater than 1 year, R is in the middle value range, F and M are moderate; the decline period and low value can be defined as a significant decrease in transaction frequency and amount, and a large R value.

[0074] Step 6.2: Label each type of customer.

[0075] Based on the previous cross-combination results, define feature labels for each customer group. For example, basic information labels, such as demographic characteristics such as age, gender, and region; behavioral feature labels, such as transaction frequency, account balance fluctuations, product usage preferences, etc.; life cycle labels, such as which life cycle stage the customer is currently in (introduction, growth, maturity, decline); value labels: whether the customer is a high-value, medium-value, or low-value customer; behavioral trend labels: such as whether the transaction behavior is growing, stable, or declining. For example, in the growth stage and high value, its label can be "high transaction frequency, increased activity, preference for using financial products, and possible interest in loan products"; in the decline stage and low value, its label may be "decreased transaction frequency, long-term non-use of banking products, need to be awakened."

[0076] The same applies to new customers.

[0077] Step 7: Divide the product value levels according to the product data to obtain product value stratification.

[0078] The value level is divided according to the direct income generated by the product, such as the loan interest rate of the loan product, or the indirect income generated by the deposit product. Products can be divided into high-value products, medium-value products and low-value products. For example, products that can bring high income to the bank are classified as high-value products; products that can only bring limited income to the bank but can maintain existing customers are classified as medium-value products; products that basically do not bring income to the bank and are not designed to maintain existing customers are classified as low-value products.

[0079] Step 8: Conduct acceptance analysis based on customer transaction data to obtain product acceptance.

[0080] It is obtained based on the holding time and holding volume of the product in the customer. It can be divided into high acceptance, medium acceptance and low acceptance. For example, products with long holding time and high holding volume are classified as high acceptance products, products with short holding time but high holding volume are classified as medium acceptance products, and the rest are classified as low acceptance products.

[0081] Step 9: Create product profiles based on product value stratification and product acceptance, and obtain product feature labels corresponding to each type of product.

[0082] Step 9.1: Using product data and product acceptance, classify through clustering algorithm.

[0083] Step 9.2: Add corresponding feature labels to each type of feature data.

[0084] For example, a product with high acceptance and high value can be labeled as "newly listed deposit / loan product, product with a large number of purchases recently, product with a large inventory, etc."

[0085] Step 10: The existing customer personal basic data, existing customer transaction data, existing customer customer group information, existing customer cycle value and existing customer portrait constitute the existing customer feature label, and establish the first association relationship between the existing customer feature label and the product feature label of the product; the new customer personal basic data, new customer transaction data, new customer customer group information, new customer cycle value and new customer portrait constitute the new customer feature label, and establish the second association relationship between the new customer feature label and the product feature label of the product.

[0086] The process of establishing the first association includes the recall stage, the sorting stage, and the rearrangement stage. In the recall stage, a large number of customers in the bank are quickly and efficiently positioned for products. The main algorithms are: collaborative filtering based on customers (UserCF) and products (ItemCF), SVD and other algorithms; in the sorting stage, machine learning or deep learning models are used to further deeply mine the results of the recall stage, including customer and product portrait features mined in the previous stage, as well as interactive features, etc. The main algorithms are GBDT+LR fusion model, DNN model, DeepFM model, etc.; in the rearrangement stage, the results of the recall stage and the sorting stage are re-filtered, sorted, and filtered and output on the business side, such as filtering of products on sale or that can be recommended, filtering of recommended target customers, and shuffling or sorting the recommended results according to other rules.

[0087] In the process of establishing the second association relationship, a collaborative filtering model is used for the new customer recommendation model, including calculating cosine similarity, distance calculation model and similarity calculation model; it is also divided into product-based collaborative filtering and customer-based collaborative filtering.

[0088] After step 10, a personalized recommendation list for existing / new customers is output, which includes the expected purchase probability of each customer for each product, and the top N products are selected based on the probability.

[0089] Step 11: Output product recommendation results for existing customers based on the first association relationship, and output product recommendation results for new customers based on the second association relationship.

[0090] Step 12: Results display, including customer label data and customer profile data output by the model.

[0091] Corresponding to the above-mentioned first aspect, the second aspect of this embodiment provides a banking financial product recommendation system, including:

[0092] The collection module is used to collect basic personal data of customers, customer transaction data and product data;

[0093] A classification module, used to classify customer personal basic data into existing customer personal basic data and new customer personal basic data, and to classify customer transaction data into existing customer transaction data and new customer transaction data;

[0094] The customer group positioning module is used to divide existing customers into different customer groups to obtain the customer groups to which existing customers belong; it is used to divide new customers into different customer groups to obtain the customer groups to which new customers belong;

[0095] The cycle value module is used to perform life cycle analysis and value stratification on existing customers to obtain the cycle value curve of existing customers; it is used to perform life cycle analysis and value stratification on new customers to obtain the cycle value curve of new customers;

[0096] The customer positioning and profiling module is used to perform high-dimensional cross-precision positioning of the status of existing customers based on their personal basic data, customer group information and cycle value, and to perform customer profiling to obtain the customer portrait corresponding to each type of existing customer group; perform high-dimensional cross-precision positioning of the status of new customers based on their personal basic data, customer group information and cycle value, and to perform customer profiling to obtain the customer portrait corresponding to each type of new customer group;

[0097] The product portrait module is used to divide the product value levels according to product data to obtain product value stratification; to analyze the acceptance according to customer transaction data to obtain product acceptance; to perform product portraits according to product value stratification and product acceptance to obtain product feature labels corresponding to each type of product;

[0098] An association module is used to construct a feature label of an existing customer based on the basic personal data of the existing customer, the customer group information of the existing customer, the cycle value and the customer portrait; to construct a feature label of a new customer based on the basic personal data of the new customer, the customer group information of the new customer, the cycle value and the customer portrait; to establish a first association relationship between the feature label of the existing customer and the product feature label of the product, and to establish a second association relationship between the feature label of the new customer and the product feature label of the product;

[0099] The recommendation module is used to output product recommendation results for existing customers according to the first association relationship, including product recommendations for existing customers, customer service recommendations for existing customers, and message recommendations for existing customers; and to output product recommendation results for new customers according to the second association relationship, including product recommendations for new customers, customer service recommendations for new customers, and message recommendations for new customers.

[0100] The data processing module is used to pre-process the collected customer personal basic data, customer transaction data and product data.

[0101] Among them, the data processing module includes: a data cleaning unit for cleaning invalid data and duplicate data; a standardization processing unit for standardizing the cleaned data. The customer group positioning module includes: a customer group classification unit for classifying customer groups using a clustering algorithm; a model conversion module for converting customers in different customer life cycles. The cycle value module includes: a life cycle positioning module for dividing the life cycle of existing customers according to the personal basic data of existing customers and the transaction data of existing customers to obtain the life cycle of existing customers; dividing the life cycle of new customers according to the personal basic data of new customers and the transaction data of new customers to obtain the life cycle of new customers. The life cycle positioning module includes: a cycle division module, which is used to divide the customer life cycle into an introduction period, a growth period, a mature period and a decline period; a customer group division module, which is used to divide customers into multiple customer groups, including introduction period customers, growth period customers, mature period customers and decline period customers, based on the collected customer personal basic data characteristics, customer transaction data characteristics and customer life cycle characteristics; a value stratification positioning module, which is used to perform value stratification of existing customers based on the personal basic data of existing customers and the transaction data of existing customers to obtain the corresponding value stratification of existing customers; and to perform value stratification of new customers based on the personal basic data of new customers and the transaction data of new customers to obtain the corresponding value stratification of new customers. The value stratification and positioning module includes: a value stratification module, which is used to divide the customer value level into high value, medium value and low value according to the RFM model; a customer group division module, which is used to divide customers into multiple customer groups, including high-value customers, medium-value customers and low-value customers, according to the collected customer personal basic data characteristics, customer transaction data characteristics and customer value stratification characteristics; a life cycle curve drawing module, which is used to cross-analyze the existing customer life cycle value according to the existing customer life cycle and existing customer value stratification, and obtain the corresponding cycle value of each type of existing customer group; cross-analyze the new customer life cycle value according to the new customer life cycle and new customer value stratification, and obtain the corresponding cycle value of each type of new customer group; the customer positioning portrait module includes: a coordinate construction unit, which performs high-dimensional cross-precision positioning of the existing customer status according to the existing customer personal basic data, the existing customer group information and the cycle value, and obtains the customer status coordinates; a data classification unit, which is used to classify the customer status coordinates using a clustering algorithm; a label adding unit, which is used to add corresponding feature labels to each type of feature data. The product portrait module includes: an acceptance analysis unit, which is used to use customer transaction data to model product transaction status and obtain product acceptance; a value stratification unit, which is used to use product data to perform product value stratification and obtain product value stratification; a data classification unit, which is used to use product value stratification and product acceptance to perform classification through a clustering algorithm; and a label adding unit, which is used to add corresponding feature labels to each type of feature data.

[0102] Based on the method provided in the first aspect and the system provided in the second aspect, the third aspect of this embodiment provides a computer device for executing the method described in the first aspect or any method that may be involved in the method described in the first aspect, including a memory, a processor and a transceiver that are sequentially connected in communication, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the method described in the first aspect or any method that may be involved in the method described in the first aspect. For example, the memory may include, but is not limited to, a random access memory (Random-Access Memory, RAM), a read-only memory (Read-Only Memory, ROM), a flash memory (FlashMemory), a first-in-first-out memory (First Input First Output, FIFO) and / or a first-in-last-out memory (First Input Last Output, FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may also include, but is not limited to, a power module, a display screen and other necessary components.

[0103] The working process, working details and technical effects of the aforementioned computer device provided in the third aspect of this embodiment can be referred to the method described in the first aspect or any method that may be involved in the first aspect, and will not be described in detail here.

[0104] The fourth aspect of this embodiment provides a computer-readable storage medium that stores the method described in the first aspect or any method that may be related to the method described in the first aspect, that is, the computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the method described in the first aspect or any method that may be related to the method described in the first aspect is executed. The computer-readable storage medium refers to a carrier for storing data, which may include but is not limited to computer-readable storage media such as floppy disks, optical disks, hard disks, flash memories, USB flash drives, and / or memory sticks, and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0105] The working process, working details and technical effects of the aforementioned computer-readable storage medium provided in the fourth aspect of this embodiment can be referred to the method described in the first aspect or any method that may be related to the method described in the first aspect, and will not be repeated here.

[0106] In a fifth aspect, the present embodiment provides a computer program product comprising instructions, which, when executed on a computer, causes the computer to execute the method described in the first aspect or any method that may be related to the method described in the first aspect. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0107] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A banking financial product recommendation method, characterized in that: The following steps are involved: S1: Collect basic personal data, transaction data and product data of customers; S2: Divide the customer's personal basic data into the existing customer's personal basic data and the new customer's personal basic data; S3: Divide the customer transaction data into existing customer transaction data and new customer transaction data; S4: Based on the basic personal data of existing customers, the customer group information of existing customers and the cycle value of existing customers, the status of existing customers is accurately positioned by high-dimensional cross-pollination, and customer profiling is performed to obtain the customer profile corresponding to each type of existing customer group; based on the basic personal data of new customers, the customer group information of new customers and the cycle value of new customers, the status of new customers is accurately positioned by high-dimensional cross-pollination, and customer profiling is performed to obtain the customer profile corresponding to each type of new customer group; product profiling is performed based on product value stratification and product acceptance to obtain the product feature labels corresponding to each type of product; S5: Construct feature labels for existing customers based on their personal basic data, customer group information, cycle value and customer portraits; construct feature labels for new customers based on their personal basic data, customer group information, cycle value and customer portraits; S6: establishing a first association relationship between the existing customer feature tag and the product feature tag of the product, and establishing a second association relationship between the new customer feature tag and the product feature tag of the product; S7: Outputting product recommendation results for existing customers according to the first association relationship, including existing customer product recommendations, existing customer service recommendations, and existing customer message recommendations; Outputting product recommendation results for new customers according to the second association relationship, including new customer product recommendations, new customer service recommendations, and new customer message recommendations; S5 includes: S5.1: According to the life cycle theory, the customer life cycle is divided into the introduction stage, growth stage, maturity stage and decline stage; S5.2: Based on the collected basic personal data characteristics, customer transaction data characteristics and customer life cycle characteristics, customers are divided into multiple customer groups, including introduction customers, growth customers, mature customers and decline customers; S5.3: Identify when a customer transitions from one lifecycle stage to another by regularly analyzing customer behavior data; trigger the product recommendation model to transition when a customer enters a new lifecycle stage; define the transition conditions for each stage based on account activity, transaction frequency, and product usage; After S1, the following steps are also included: preprocessing the collected basic personal data of customers, customer transaction data and product data; the data preprocessing includes: cleaning invalid data and duplicate data; and standardizing the cleaned data; After S3 and before S4, the following steps are also included: Perform customer group attribution analysis based on the personal basic data and transaction data of existing customers to obtain customer group information of existing customers; perform customer group attribution analysis based on the personal basic data and transaction data of new customers to obtain customer group information of new customers; The existing customer life cycle is positioned based on the customer group affiliation, the existing personal basic data and the existing customer transaction data to obtain the existing customer life cycle; the new customer life cycle is positioned based on the customer group affiliation, the new customer personal basic data and the new customer transaction data to obtain the new customer life cycle; The existing customers are valued according to the customer group affiliation, the existing personal basic data and the existing customer transaction data, and the corresponding value stratification of the existing customers is obtained; the new customers are valued according to the customer group affiliation, the new customers' personal basic data and the new customers' transaction data, and the corresponding value stratification of the new customers is obtained; According to the existing customer life cycle and the existing customer value stratification, the cycle value of existing customers is cross-analyzed to obtain the cycle value corresponding to each type of existing customer group; according to the new customer life cycle and the new customer value stratification, the cycle value of new customers is cross-analyzed to obtain the cycle value corresponding to each type of new customer group; Divide product value levels according to product data to obtain product value stratification; conduct acceptance analysis based on customer transaction data to obtain product acceptance.

2. A banking financial product recommendation method according to claim 1, characterized in that: Customer lifecycle targeting includes the following steps: Divide the customer life cycle into the introduction stage, growth stage, maturity stage and decline stage; Divide customers into multiple customer groups based on their personal basic data, customer transaction data and customer life cycle, including introduction customers, growth customers, mature customers and decline customers; Transform product recommendation models at different customer life cycles.

3. A banking financial product recommendation method according to claim 1, characterized in that: Customer value stratification includes the following steps: Use the RFM model to stratify customer value; Divide customer value into high value, medium value and low value; Based on customer personal basic data, customer transaction data and customer value stratification, customers are divided into multiple customer groups, including high-value customers, medium-value customers and low-value customers.

4. A banking financial product recommendation method according to claim 1, characterized in that: Conducting customer cross-analysis involves the following steps: The obtained customer life cycle and customer value stratification are used to build a model and obtain the customer life cycle value curve.

5. A banking financial product recommendation method according to claim 1, characterized in that: The following steps are involved in creating a customer profile: Using basic customer information, personal transaction data, customer group data and cycle value curve, we conduct multi-dimensional analysis of customers and obtain customer status coordinates; Use clustering algorithms to classify customer status coordinates; Add corresponding feature labels to each type of feature data.

6. A banking financial product recommendation method according to claim 1, characterized in that: The following steps are involved in conducting a product portrait: Use customer transaction data to model product transaction status and obtain product acceptance; Use product data to model product value and obtain product value stratification; Using product data and product acceptance, classification is performed through clustering algorithms; Add corresponding feature labels to each type of data.

7. A computer device, characterized in that: It comprises a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive data, and the processor is used to read the computer program to execute the bank financial product recommendation method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the bank financial product recommendation method according to any one of claims 1 to 6 is executed.

Images