CN113793061A - Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM - Google Patents
Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM Download PDFInfo
- Publication number
- CN113793061A CN113793061A CN202111132743.0A CN202111132743A CN113793061A CN 113793061 A CN113793061 A CN 113793061A CN 202111132743 A CN202111132743 A CN 202111132743A CN 113793061 A CN113793061 A CN 113793061A
- Authority
- CN
- China
- Prior art keywords
- user
- data
- layer
- rfm
- rating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/242—Query formulation
- G06F16/2433—Query languages
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/02—Banking, e.g. interest calculation or account maintenance
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Databases & Information Systems (AREA)
- Educational Administration (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Accounting & Taxation (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Technology Law (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of informatization, and provides a business bank customer rating method and device integrating an analytic hierarchy process (RFM) and an RFM, aiming at solving the problems of poor model rating effect caused by colinearity, single index and mismatching of actual business of RFM model variables in deposit business. The method comprises the main implementation steps of determining a target layer and a criterion layer, wherein the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value; collecting user data sets of three aspects of user activity, user loyalty and user value; performing data cleaning on the related data; respectively constructing scheme layer indexes of three criteria of user activity, user value and user loyalty by using an RFM (remote sensing and mapping) model; confirming scheme layer index weight based on an analytic hierarchy process, outputting a criterion layer index result, determining the criterion layer index weight, and outputting a grading rating of a target layer client; and finishing the formulation and popularization of the differential marketing strategy or activity based on the grading rating of the target layer client.
Description
Technical Field
The invention relates to the technical field of informatization, and provides a business bank customer rating method and device integrating an analytic hierarchy process (analytic hierarchy process) and a Radio Frequency Memory (RFM).
Background
With the development of social economy and the wide application of mobile internet, the scale of the behavior data of commercial banks is increased, and a problem of how to discover the user value from the huge user behavior data is solved; on the other hand, the strengthening of financial supervision, the homogenization of products among commercial banks and the aggravation of industry competition put higher demands on the level of operation of commercial bank customers and the control of marketing cost. Therefore, based on the user behavior data, the business bank customers are well graded, and the accurate, individual and low-cost marketing campaign is very important.
The traditional RFM (Recency, Frequency and money) model mainly uses three elements of recent consumption (Recency), consumption Frequency (Frequency) and consumption amount (money), and respectively reflects three aspects of the loss possibility of a user, the loyalty of the user and the value of the user, divides the user into three-dimensional quadrants through three indexes, and adopts different marketing strategies for the user based on the quadrant position of the user. Because the critical value for distinguishing the high and low of a user in a certain dimension has certain subjectivity and is often difficult to determine, the common mode is to cluster three indexes by using a k-means clustering mode and explain each class through the position of the mass center. Generally, customers are divided into several levels, namely, worthless customers, general customers, important reserve customers, general important customers, important maintenance customers and important development customers, and different marketing strategies are adopted for the customers based on different user levels.
The RFM model displays the entire outline of a client more dynamically, which provides a basis for personalized communication and service, but has many disadvantages. Firstly, the critical value of a traditional RFM model for distinguishing users has subjectivity, and RFM using k-means clustering can avoid the subjectivity, but the classification automatically generated based on the sample distance is often susceptible to the influence of data distribution, and sometimes the classification is difficult to explain; secondly, the traditional RFM model adopts too single indexes for measuring the loss possibility of the user, the loyalty of the user and the value of the user, so that the user can not be measured fully and correctly, and divided users are rough; thirdly, the loan saving business of the commercial bank is different from the retail consumption industry, and the traditional RFM index is difficult to be applied. For example, in the user transaction frequency index for measuring the loyalty of the user, in the deposit service, the high deposit frequency of the user may be only because the deposited period is short or the same fund is repeatedly operated, and in the index for measuring the value of the user, the higher the transaction frequency of the user is, the larger the total transaction amount is, the higher the user value is, but the actual situation may be the opposite; in addition, the RFM model of the deposit performs equal-weight processing on three different indexes, and under the actual business scene, the importance of each variable may be different, and the co-linearity often exists among the variables, which also affects the model effect.
Disclosure of Invention
Based on the foregoing, the present invention improves upon the conventional RFM model to address many of the deficiencies of conventional RFM in user ratings. We solve the above technical problem using a bank customer rating method that combines models of analytic hierarchy process and RFM. In the selection of the indexes, a plurality of indexes are selected to represent user characteristics of three dimensions of user activity, user loyalty and user value. In a grading manner, the user score is calculated by using an analytic hierarchy process for grading.
In order to achieve the purpose, the invention adopts the following technical means:
the invention provides a business bank customer rating method combining an analytic hierarchy process and RFM, which comprises the following implementation processes:
step 1, determining a target layer and a criterion layer, wherein the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value;
step 2, collecting user data sets of three aspects of user activity, user loyalty and user value;
step 3, data cleaning is carried out on the data related to the step 2;
step 4, respectively constructing scheme layer indexes of three criteria of user activity, user value and user loyalty by using an RFM model; confirming scheme layer index weight based on an analytic hierarchy process, outputting a criterion layer index result, determining the criterion layer index weight, and outputting a grading rating of a target layer client;
and 5, finishing the formulation and popularization of the differential marketing strategy or activity based on the grading rating of the target layer client.
In the above technical solution, the user data set used includes four parts:
the first part is basic attribute data of the user;
the second part is user behavior data which is trace data in the process of operating the APP by the user;
the third part is user transaction data, including payment data of the user's off-line scene, data of user's transaction deposit and loan products;
the fourth part is that the user redeems the points for the recorded data.
Optionally, the selection of the user activity index can select the user login times, the continuous login days of the user, the time length from the last time the user holds a deposit product to the present, the time length from the present time the user opens an account, the times of browsing the product page by the user, the stay time of the product page of the user, the times of using the APP mini-game by the user, the times of exchanging the gift by the user using the credit, and the like; the selection of the loyalty index of the user can select the number of new users shared by the user, the frequency of product purchase of the user, the frequency of point exchange of the user, the number of days of deposit holding of the user and the like; the user value index can be selected from the monthly and daily average deposit balance of the user, the historical maximum single-day balance of the user and other indexes (such as city, region, age, academic calendar, occupation, house property information, vehicle information, loan data and the like) capable of reflecting the property condition of the user.
In the above technical solution, step 3 is executed:
step 3.1, acquiring all required data by using an SQL query statement, and loading the data by using Python;
step 3.2, carrying out variable exploration on the data, analyzing statistical values of the variables, carrying out missing value filling on the variables by combining the service characteristics of the variables, and deleting repeated data;
3.3, constructing the characteristics required by the model, including the numerical processing of the classification variables, the discretization processing of continuous variables, and the derivation of new variables which accord with business logic through variable intersection and synthesis;
optionally, smoothing (including log function transformation, Box-Cox transformation, etc.) is performed on the variables of the long-tail distribution.
And 3.4, carrying out dimensionless processing on the data obtained in the step 3.3 to obtain standardized data, wherein the selectable method comprises dimensionless processing and physicochemical dimensionless processing which have practical significance.
In the above technical solution, step 4 is executed to establish a hierarchical analysis model for the processed data, and the specific operation content includes:
step 4.1, constructing a hierarchical structure model based on the standardized data obtained in the step 3.4, wherein the structure comprises a target layer, a criterion layer and a scheme layer; the target layer is the grading rating of the user, the criterion layer is divided into three aspects of user activity, user value and user loyalty, and the scheme layer is an index for further refining the index of the rule layer;
4.2, comparing the importance of the subdivision variables under each criterion layer to construct a pair comparison matrix for judging the importance degree of the variables;
and 4.3, carrying out consistency check on the formed comparison matrix: and calculating the maximum characteristic value and the corresponding characteristic vector of each paired comparison matrix, and carrying out consistency check, wherein if the maximum characteristic value and the corresponding characteristic vector pass the consistency check, the normalized characteristic vector is a weight vector of a variable, otherwise, the maximum characteristic value and the corresponding characteristic vector need to be reconstructed into the comparison matrix. This process checks whether the importance conflicts between the metrics.
4.4, calculating indexes of a criterion layer, namely the activity of a user, the value of the user and the loyalty score of the user;
and 4.5, repeating the method of the step 4.2-4.4, calculating the customer rating of the target layer, and finishing user grading according to the score.
The business bank customer rating method based on the analytic hierarchy process and the RFM model is specifically represented as follows in the user rating process: classifying users into active users, inactive customers, sleeping customers and attrition customers according to the model scores of the active aspects of the users; classifying the users into potential customers, loyalty customers and decline customers according to the scores of the users in terms of loyalty; according to the model score in the aspect of user value, dividing the user into a low-value client and a high-value client; meanwhile, according to the scores of the hierarchical analysis model in the aspects of user activity, user loyalty and user value, the users are graded in multiple levels, and a differentiated marketing strategy can be carried out.
The user rating method for expanding the requirements of specific marketing scenes and marketing targets mainly comprises the following steps: layering according to a user purchasing decision mode, and classifying users into impulse purchasing type, repeated hesitation type and rational comparison type; and classifying the users into activity sensitivity type and activity insensitivity type according to the sensitivity hierarchy of the users to the marketing activities.
The invention also provides a commercial bank customer rating device integrating the analytic hierarchy process and the RFM, which comprises the following implementation processes:
the target layer and criterion layer module is used for determining a target layer and a criterion layer, the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value;
the system comprises a user data set module, a data processing module and a data processing module, wherein the user data set module is used for collecting user data sets in three aspects of user activity, user loyalty and user value;
the cleaning module is used for cleaning data in the user data set module;
the grading module is used for respectively constructing scheme layer indexes of three criteria of user activity, user value and user loyalty by utilizing an RFM (remote sensing module); confirming scheme layer index weight based on an analytic hierarchy process, outputting a criterion layer index result, determining the criterion layer index weight, and outputting a grading rating of a target layer client;
and the marketing module finishes the formulation and popularization of the differential marketing strategy or activity based on the grading rating of the target layer customers.
In the above technical solution, the user data set used includes four parts:
the first part is basic attribute data of the user;
the second part is user behavior data which is trace data in the process of operating the APP by the user;
the third part is user transaction data, including payment data of the user's off-line scene, data of user's transaction deposit and loan products;
the fourth part is that the user redeems the points for the recorded data.
In the above technical scheme, in the cleaning module:
step 1, acquiring all required data by using an SQL query statement, and loading the data by using Python;
step 2, carrying out variable exploration on the data, analyzing statistical values of variables, carrying out missing value filling on the variables by combining the service characteristics of the variables, and deleting repeated data;
step 3, constructing the characteristics required by the model, including the numerical processing of the classification variables, the discretization processing of continuous variables, and the derivation of new variables which accord with business logic through variable intersection and synthesis;
optionally, smoothing (including log function transformation, Box-Cox transformation, etc.) is performed on the variables of the long-tail distribution.
And 4, carrying out non-dimensionalization treatment on the data obtained in the step 3 to obtain standardized data, wherein the selectable method comprises dimension treatment with practical significance and dimension treatment with physicochemical property.
In the above technical solution, the executing rating module establishes a hierarchical analysis model for the processed data, and the specific operation content includes:
step 1, constructing a hierarchical analysis model based on standardized data obtained by a cleaning module, wherein the structure comprises a target layer, a criterion layer and a scheme layer; the target layer is the grading rating of the user, the criterion layer is divided into three aspects of user activity, user value and user loyalty, and the scheme layer is an index for further refining the index of the rule layer;
step 2, comparing the importance of the subdivision variables under each criterion layer to construct a paired comparison matrix for judging the importance degree of the variables;
step 3, carrying out consistency check on the formed comparison matrix: and calculating the maximum characteristic value and the corresponding characteristic vector of each paired comparison matrix, and carrying out consistency check, wherein if the maximum characteristic value and the corresponding characteristic vector pass the consistency check, the normalized characteristic vector is a weight vector of a variable, otherwise, the maximum characteristic value and the corresponding characteristic vector need to be reconstructed into the comparison matrix. This process checks whether the importance conflicts between the metrics.
Step 4, calculating indexes of a criterion layer, such as the activity of a user, the value of the user and the loyalty score of the user;
and 5, repeating the methods in the steps 2-4, calculating the customer rating of the target layer, and finishing the user grading according to the score.
The invention also provides a storage medium, which stores a program for the commercial bank customer rating by fusing the analytic hierarchy process and the RFM, and a processor executes the program to realize the commercial bank customer rating method by fusing the analytic hierarchy process and the RFM.
Because the invention adopts the technical scheme, the invention has the following beneficial effects:
in the aspect of variable selection, the invention screens the user characteristics of multiple dimensions respectively aiming at three aspects of user loss possibility, user loyalty and user value in the RFM model. The model features are richer, the description on the activity of the user, the loyalty of the user and the value of the user is more refined, and the model more meets the actual business requirements of commercial banks.
In a hierarchical manner, the user composite score is calculated by using an analytic hierarchy process, and logical experience, insight and intuition are introduced. The subjectivity of the traditional layering method is avoided to a certain extent. Meanwhile, different weights are used according to different importance of different variables, the method is more in line with business reality, and the grading result is more scientific and explanatory.
In the expansibility of the model, a new user rating model can be constructed aiming at a specific marketing scene and a marketing target, so that the rating is richer compared with the traditional RFM model. The expanded hierarchical model can be cross-hierarchically graded with the RFM model, and the feature depiction of the client is more three-dimensional and richer.
Drawings
FIG. 1 is a schematic flow chart of a method and apparatus for commercial bank customer rating incorporating analytic hierarchy process and RFM according to the present invention;
FIG. 2 is a schematic diagram of a hierarchical structure of a business bank customer rating method and apparatus combining an analytic hierarchy process and RFM according to the present invention.
Detailed Description
The invention will be further described with reference to the accompanying drawings and specific embodiments.
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Step 1, determining a target layer and a criterion layer, wherein the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value.
And 2, collecting user data sets of the aspects of user activity, user loyalty and user value. The related user behavior data set comprises four parts, namely basic attribute data of a user, user behavior data, user transaction data and point use redemption record data of the user.
And 3, selecting specific indexes for three different decision targets. Selecting user activity indexes, namely selecting user login times, user continuous login days, time from the last time that a user holds a deposit product to the present, time from the present when the user opens an account, times that the user browses a product page, time that the user stays on the product page, times that the user uses an APP mini-game, times that the user uses credits to exchange gifts and the like; the selection of the loyalty index of the user can select the number of new users shared by the user, the frequency of product purchase of the user, the frequency of point exchange of the user, the number of days of deposit holding of the user and the like; the user value index can be selected from the monthly and daily average deposit balance of the user, the historical maximum single-day balance of the user and other indexes (such as city, region, age, academic calendar, occupation, house property information, vehicle information, loan data and the like) capable of reflecting the property condition of the user.
Step 4, cleaning the user data set, sequentially acquiring all required data by using SQL query statements and loading the data by using Python; carrying out variable exploration on data, analyzing statistical values of variables, carrying out missing value filling on the variables by combining the service characteristics of the variables, and deleting repeated data; characteristics required for constructing the model: the method comprises the steps of carrying out numerical processing on classified variables, carrying out discretization processing on continuous variables, and deriving new variables which accord with business logic through variable crossing and synthesis; smoothing variables distributed by the long tail (including log function transformation, Box-Cox transformation and the like); the data is subjected to dimensionless processing to obtain standardized data, and the optional method comprises dimensionless processing and physicochemical dimensionless processing which have practical significance.
Step 5, establishing a hierarchical analysis model for the processed data, wherein the specific operation contents comprise: and (4) constructing a hierarchical structure model based on the data obtained in the step (4), wherein the structure comprises a target layer, a criterion layer and a scheme layer. The target layer is the grading rating of the user, the criterion layer is divided into three aspects of user activity, user value and user loyalty, and the scheme layer is an index for further refining the index of the rule layer; comparing the importance of the subdivision variables under each criterion layer to construct a pair comparison matrix for judging the importance degree of the variables; and (3) carrying out consistency check on the comparison matrix: and calculating the maximum characteristic value and the corresponding characteristic vector of each paired comparison matrix, and carrying out consistency check, wherein if the maximum characteristic value and the corresponding characteristic vector pass the consistency check, the normalized characteristic vector is the weight vector of the variable, otherwise, the comparison matrix needs to be reconstructed. This process checks whether the importance conflicts between the metrics. And calculating the indexes of the criterion layer, calculating the scores of user activity, user value and user loyalty, and finishing user grading according to the scores.
Step 6, dividing the users into active users, inactive customers, sleeping customers and attrition customers according to the model scores of the active aspects of the users; classifying the users into potential customers, loyalty customers and decline customers according to the scores of the users in terms of loyalty; according to the model score in the aspect of user value, dividing the user into a low-value client and a high-value client; meanwhile, according to the score of the hierarchical analysis model in R, F, M, the users are graded in multiple levels, and a differentiated marketing strategy can be performed. The user rating method for expanding the requirements of specific marketing scenes and marketing targets mainly comprises the following steps: layering according to a user purchasing decision mode, and classifying users into impulse purchasing type, repeated hesitation type and rational comparison type; and classifying the users into activity sensitivity type and activity insensitivity type according to the sensitivity hierarchy of the users to the marketing activities.
And 7, compiling the modeling process into an automatic scheduled program, enabling the program to be automatically executed in a rating device, and enabling the rating device to automatically output the rating and classification of each user to realize the division of the customer groups.
And 8, carrying out differentiated marketing on different customer groups. In the application of the RFM model, based on the clients with actively layered users, some interesting mini games or welfare activities are pushed to the clients with low activity, some new products or new functions can be pushed to the active users to enhance the experience, and the sleeping clients are required to be given stronger reward strength to avoid losing; based on customers of user loyalty stratification, potential customers and new customers mainly develop habits and stickiness of purchasing products, and for loyal customers, mining user value is mainly used; based on the customers of user value layering, can improve marketing efficiency according to the differentiated marketing cost of different customer value distribution, more combine other layering models, formulate the strategy of differentiation: for example, sleeping customers are often difficult to recover and should not pay greater marketing costs in principle, while for high value, losing customers, it is possible to reach and push back activities appropriately; in the expanded user rating model, based on users who buy the decision-making mode layering, the motivation type customer can increase the reward intensity of the present activities, repeatedly stimulate, and the motivation type customer is rationally compared, can launch the product with more cost performance than other banks in the product design, and the customer of the repeated hesitation type should avoid its repeated transaction, can adopt the scheme of issuing the activity reward by stages, and guarantee its retention.
Claims (9)
1. A method for rating customers of a commercial bank by combining an analytic hierarchy process and an RFM, the method comprising:
step 1, determining a target layer and a criterion layer, wherein the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value;
step 2, collecting user data sets of three aspects of user activity, user loyalty and user value;
step 3, data cleaning is carried out on the data related to the step 2;
step 4, respectively constructing scheme layer indexes of three criteria of user activity, user value and user loyalty by using an RFM model; confirming scheme layer index weight based on an analytic hierarchy process, outputting a criterion layer index result, determining the criterion layer index weight, and outputting a grading rating of a target layer client;
and 5, finishing the formulation and popularization of the differential marketing strategy or activity based on the grading rating of the target layer client.
2. The method of integrating analytic hierarchy process and RFM for business bank customer rating according to claim 1, wherein: the user data set concerned comprises four parts:
the first part is basic attribute data of the user;
the second part is user behavior data which is trace data in the process of operating the APP by the user;
the third part is user transaction data, including payment data of the user's off-line scene, data of user's transaction deposit and loan products;
the fourth part is the record data of the user for exchanging the points for the commodities.
3. The method of integrating analytic hierarchy process and RFM for business bank customer rating according to claim 1, wherein: and (3) executing the step of data cleaning:
step 3.1, acquiring all required data by using an SQL query statement, and loading the data by using Python;
step 3.2, carrying out variable exploration on the data, analyzing statistical values of the variables, carrying out missing value filling on the variables by combining the service characteristics of the variables, and deleting repeated data;
3.3, constructing the characteristics required by the model, including the numerical processing of the classification variables, the discretization processing of continuous variables, and the derivation of new variables which accord with business logic through variable intersection and synthesis;
and 3.4, carrying out dimensionless processing on the data obtained in the step 3.3 to obtain standardized data, wherein the selectable method comprises dimensionless processing and physicochemical dimensionless processing which have practical significance.
4. The method of integrating analytic hierarchy process and RFM for business bank customer rating according to claim 1, wherein: and 4, executing step 4, establishing a hierarchical analysis model for the processed data, wherein the specific operation contents comprise:
step 4.1, constructing a hierarchical analysis model based on the standardized data obtained in the step 3.4, wherein the structure comprises a target layer, a criterion layer and a scheme layer; the target layer is the grading rating of the user, the criterion layer is divided into three aspects of user activity, user value and user loyalty, and the scheme layer is an index for further refining the index of the rule layer;
4.2, carrying out pairwise comparison on the importance of the subdivision variables under each criterion layer to construct a pairwise comparison matrix for judging the importance degree of the variables;
and 4.3, carrying out consistency check on the formed comparison matrix: calculating the maximum eigenvalue and the corresponding eigenvector of each paired comparison matrix, and performing consistency check, wherein if the maximum eigenvalue and the corresponding eigenvector pass the consistency check, the normalized eigenvector is a weight vector of a variable, otherwise, the paired comparison matrix needs to be reconstructed;
4.4, calculating indexes of a criterion layer, namely the activity of a user, the value of the user and the loyalty score of the user;
and 4.5, repeating the method of the step 4.2-4.4, calculating the customer rating of the target layer, and finishing user grading according to the score.
5. A device for integrating analytic hierarchy process (analytic hierarchy process) and RFM for rating customers of commercial banks, which is characterized in that the implementation process comprises the following steps:
the target layer and criterion layer module is used for determining a target layer and a criterion layer, the target layer is used for grading the user, and the criterion layer is divided into three aspects of user activity, user loyalty and user value;
the system comprises a user data set module, a data processing module and a data processing module, wherein the user data set module is used for collecting user data sets in three aspects of user activity, user loyalty and user value;
the cleaning module is used for cleaning data in the user data set module;
the grading module is used for respectively constructing scheme layer indexes of three criteria of user activity, user value and user loyalty by utilizing an RFM (remote sensing module); confirming scheme layer index weight based on an analytic hierarchy process, outputting a criterion layer index result, determining the criterion layer index weight, and outputting a grading rating of a target layer client;
and the marketing module finishes the formulation and popularization of the differential marketing strategy or activity based on the grading rating of the target layer customers.
6. The fused analytic hierarchy process and RFM commercial bank customer rating device of claim 5, wherein: the user data set concerned comprises four parts:
the first part is basic attribute data of the user;
the second part is user behavior data which is trace data in the process of operating the APP by the user;
the third part is user transaction data, including payment data of the user's off-line scene, data of user's transaction deposit and loan products;
the fourth part is that the user redeems the points for the recorded data.
7. The fused analytic hierarchy process and RFM commercial bank customer rating device of claim 5, wherein: in the cleaning module:
step 1, acquiring all required data by using an SQL query statement, and loading the data by using Python;
step 2, carrying out variable exploration on the data, analyzing statistical values of variables, carrying out missing value filling on the variables by combining the service characteristics of the variables, and deleting repeated data;
step 3, constructing the characteristics required by the model, including the numerical processing of the classification variables, the discretization processing of continuous variables, and the derivation of new variables which accord with business logic through variable intersection and synthesis;
optionally, smoothing is performed on the variable of the long-tail distribution.
And 4, carrying out dimensionless processing on the data obtained in the step 3 to obtain standardized data, wherein the selectable method comprises dimensionless processing and physicochemical dimensionless processing which have practical significance.
8. The fused analytic hierarchy process and RFM commercial bank customer rating device of claim 5, wherein: and the execution rating module is used for establishing a hierarchical analysis model for the processed data, and the specific operation contents comprise:
step 1, constructing a hierarchical analysis model based on standardized data obtained in a data cleaning module, wherein the structure comprises a target layer, a criterion layer and a scheme layer; the target layer is the grading rating of the user, the criterion layer is divided into three aspects of user activity, user value and user loyalty, and the scheme layer is an index for further refining the index of the rule layer;
step 2, comparing the importance of the subdivision variables under each criterion layer to construct a paired comparison matrix for judging the importance degree of the variables;
step 3, carrying out consistency check on the formed comparison matrix: and calculating the maximum characteristic value and the corresponding characteristic vector of each paired comparison matrix, and carrying out consistency check, wherein if the maximum characteristic value and the corresponding characteristic vector pass the consistency check, the normalized characteristic vector is a weight vector of a variable, otherwise, the maximum characteristic value and the corresponding characteristic vector need to be reconstructed into the comparison matrix. This process checks whether the importance conflicts between the metrics.
Step 4, calculating indexes of a criterion layer, such as the activity of a user, the value of the user and the loyalty score of the user;
and 5, repeating the methods in the steps 2-4, calculating the customer rating of the target layer, and finishing the user grading according to the score.
9. A storage medium storing a program for a fused analytic hierarchy process and RFM commercial bank customer rating, the program when executed by a processor implementing a fused analytic hierarchy process and RFM commercial bank customer rating method as claimed in claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111132743.0A CN113793061A (en) | 2021-09-27 | 2021-09-27 | Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111132743.0A CN113793061A (en) | 2021-09-27 | 2021-09-27 | Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113793061A true CN113793061A (en) | 2021-12-14 |
Family
ID=78877258
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111132743.0A Pending CN113793061A (en) | 2021-09-27 | 2021-09-27 | Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113793061A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115729923A (en) * | 2022-12-06 | 2023-03-03 | 广州天维信息技术股份有限公司 | Customer data analysis method and device, storage medium and computer equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391680A (en) * | 2017-07-24 | 2017-11-24 | 北京京东尚科信息技术有限公司 | Content recommendation method, device and equipment |
| CN109583651A (en) * | 2018-12-03 | 2019-04-05 | 焦点科技股份有限公司 | A kind of method and apparatus for insuring electric business platform user attrition prediction |
| CN110111156A (en) * | 2019-05-14 | 2019-08-09 | 重庆天蓬网络有限公司 | A kind of customer defection early warning method, system, medium and electronic equipment |
| CN110135901A (en) * | 2019-05-10 | 2019-08-16 | 重庆天蓬网络有限公司 | A kind of enterprise customer draws a portrait construction method, system, medium and electronic equipment |
| CN111091282A (en) * | 2019-12-10 | 2020-05-01 | 焦点科技股份有限公司 | Customer loyalty segmentation method based on user behavior data |
| CN111126775A (en) * | 2019-11-26 | 2020-05-08 | 国网浙江省电力有限公司电力科学研究院 | Hierarchical analysis method based resident customer value grading model construction method |
| US20200293923A1 (en) * | 2019-03-14 | 2020-09-17 | Punchh Inc. | Predictive rfm segmentation |
| CN112232930A (en) * | 2020-11-05 | 2021-01-15 | 西安邮电大学 | E-commerce platform customer segmentation method based on weighted RFM model |
-
2021
- 2021-09-27 CN CN202111132743.0A patent/CN113793061A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391680A (en) * | 2017-07-24 | 2017-11-24 | 北京京东尚科信息技术有限公司 | Content recommendation method, device and equipment |
| CN109583651A (en) * | 2018-12-03 | 2019-04-05 | 焦点科技股份有限公司 | A kind of method and apparatus for insuring electric business platform user attrition prediction |
| US20200293923A1 (en) * | 2019-03-14 | 2020-09-17 | Punchh Inc. | Predictive rfm segmentation |
| CN110135901A (en) * | 2019-05-10 | 2019-08-16 | 重庆天蓬网络有限公司 | A kind of enterprise customer draws a portrait construction method, system, medium and electronic equipment |
| CN110111156A (en) * | 2019-05-14 | 2019-08-09 | 重庆天蓬网络有限公司 | A kind of customer defection early warning method, system, medium and electronic equipment |
| CN111126775A (en) * | 2019-11-26 | 2020-05-08 | 国网浙江省电力有限公司电力科学研究院 | Hierarchical analysis method based resident customer value grading model construction method |
| CN111091282A (en) * | 2019-12-10 | 2020-05-01 | 焦点科技股份有限公司 | Customer loyalty segmentation method based on user behavior data |
| CN112232930A (en) * | 2020-11-05 | 2021-01-15 | 西安邮电大学 | E-commerce platform customer segmentation method based on weighted RFM model |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115729923A (en) * | 2022-12-06 | 2023-03-03 | 广州天维信息技术股份有限公司 | Customer data analysis method and device, storage medium and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Roa et al. | Super-app behavioral patterns in credit risk models: Financial, statistical and regulatory implications | |
| CN110956273A (en) | Credit scoring method and system integrating multiple machine learning models | |
| CN108256691A (en) | Refund Probabilistic Prediction Model construction method and device | |
| CN109377260A (en) | User behavior analysis system towards apparel industry | |
| US6985867B1 (en) | Method of predicting a change in an economy | |
| CN102136123A (en) | Target bank customer recognition system | |
| Chen et al. | Research on credit card default prediction based on k-means SMOTE and BP neural network | |
| CN107609771A (en) | A kind of supplier's value assessment method | |
| Salerno | Ludwig von Mises's monetary theory in light of modern monetary thought | |
| CN109213936A (en) | product search method and device | |
| JP2018503927A (en) | Segmentation and stratification of composite portfolio of investment securities | |
| Yuping et al. | New methods of customer segmentation and individual credit evaluation based on machine learning | |
| Yang et al. | The extraction of early warning features for predicting financial distress based on XGBoost model and shap framework | |
| Sun et al. | Using improved RFM model to classify consumer in big data environment | |
| CN113793061A (en) | Business bank customer rating method and device integrating analytic hierarchy process (analytic hierarchy process) and RFM | |
| Zhou et al. | A two-stage credit scoring model based on random forest: Evidence from Chinese small firms | |
| Li et al. | An empirical study on credit scoring model for credit card by using data mining technology | |
| Finlay | Towards profitability: A utility approach to the credit scoring problem | |
| CN113450004A (en) | Power credit report generation method and device, electronic equipment and readable storage medium | |
| KR20220102961A (en) | Method and apparatus of predicting default rate of individual business based on artificial intelligence model using credit information | |
| Dostál et al. | Business applications of fuzzy logic | |
| CN202093568U (en) | Bank target customer identification system | |
| Schreiber | Identification of customer groups in the German term life market: a benefit segmentation | |
| Gonçalves et al. | Credit risk analysis applying logistic regression, neural networks and genetic algorithms models | |
| Song et al. | Smart supply chain finance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |