Credit card analysis and decision-making system

1. Overview

Credit card is a high-tech retail banking business. On the one hand, the credit card (debit card) business mainly obtains profits through personal short-term recurring borrowings, which is very risky. Compared with other credit businesses, the credit card business also has the characteristics of being unsecured, having a huge number of overdrafts, and each transaction is small. characteristics, which determines that banks need to have a very complete, scientific and fast risk control method to ensure long-term and stable business development; on the other hand, with the rapid development of information technology and intensified competition, credit cards have also To become a convenient payment method, banks must provide quality services to customers in order to continue to attract and develop customers. The traditional extensive business strategy and model aimed at building a business and increasing scale has huge drawbacks and risks and is no longer suitable for the credit card business under market economy conditions. Therefore, a comprehensive, real-time and accurate credit card analysis and decision-making system covering risk control, customer relationship management, marketing and business operation analysis has become an important guarantee for banks to develop credit card business on a large scale.

As mentioned above, the analysis and decision-making of credit card business is a huge system, including many application fields and technologies, but in terms of the management level it is applicable to, it can be divided into two major categories.

The first category of applications includes basic business information query, flexible report production Adhoc (including reports and various graphics), and comprehensive business analysis, collectively referred to as business theme analysis. This kind of analysis is a summary of business data according to clear business management ideas. The technologies applied are mainly multidimensional analysis technologies (including MOLAP, ROLAP and HOLAP).

The second category is in-depth data analysis, that is, the exploration of large amounts of historical data accumulated by banks and the discovery of potential patterns. The result is often to establish a corresponding mathematical model in line with certain business goals, so that the card-issuing bank can formulate business strategies that meet market needs, including: customer credit evaluation standards, overdraft limits, annual fees, product positioning, channel selection, etc. wait. Mainly uses various data mining technologies, including classic statistical analysis and artificial intelligence modeling technology. Such as: cross frequency analysis, various linear regression analyses, non-linear regression analysis, principal component analysis, factor analysis, canonical correlation, discriminant analysis, cluster analysis, decision tree model, neuron network, etc.; Affinity Analysis, Classification analysis and factor importance analysis and other methods.

The following will explain the intelligent credit card analysis and decision-making system according to this division method.

2. Business theme analysis

Business theme analysis refers to the real-time (currently every other day, According to demand, statistics, summary and analysis can even be achieved on the same day, so that middle-level managers can grasp relevant work comprehensively, timely and accurately, and meet the various report production and analysis needs of a large number of front-line statisticians.

Business topic analysis is mainly designed to meet the needs of middle-level managers for a comprehensive and overall understanding of business conditions. Therefore, each topic analysis provides different levels of analysis based on different management perspectives. Summary data and abstract key operating indicators KPI. Not only that, the system must also allow users to obtain and share the above-mentioned business analysis data most conveniently and quickly, and be able to arbitrarily view it from multiple perspectives, including different institutions, product categories, time, currencies, etc. Carry out relevant analysis and comparison of relevant operating data (i.e. multi-dimensional analysis); at the same time, the system also provides flexible software tools to allow high-quality users to explore and research operating data according to their own management ideas.

The main subject analysis provided by the system includes:

l Card issuance volume analysis

l Transaction analysis

l Customer behavior analysis

l Customer situation analysis

l Risk analysis

l Profitability analysis

l Merchant analysis

Business theme in progress During analysis, the system provides a variety of intuitive and flexible presentation methods, including: general or customized reports, and various flexibly transformable graphics (pie charts, column charts, line charts, radar charts, and three-dimensional graphics). In terms of analysis methods, in addition to the traditional fixed base ratio and chain analysis, the system also provides the following analysis methods:

l Structural analysis

l Ratio analysis

l Trend analysis

l Principal component analysis

Among them, structural analysis and ratio analysis are performed through the system’s powerful background calculation engine Compare data from multiple different angles or levels, allowing users to gain a more comprehensive and objective grasp of business conditions; What-if analysis, trend analysis, and principal component analysis use certain mathematical models or algorithms to Analyze or fit historical data to predict and discover business conditions.

3. Decision-making prediction model

1. Credit risk control Credit Risk

For a long time, banks in my country have used manual means to conduct credit checks on newly applied customers. Reviews and ratings, and the scoring criteria used in these manual reviews basically rely on people's subjective judgment or past experience. Such credit review standards have great deviations in both accuracy and objectivity. On the other hand, for a large number of cardholders, banks lack effective means to proactively monitor and predict their credit status. As a result, they can only perform rigid checks on credit limits, transaction authorizations, overdue loans, etc. of cardholders. Extensive management. With the rapid rise of my country's credit card market, traditional credit risk control methods have increasingly exposed its flaws and problems. Credit risk has become one of the biggest obstacles to the development of my country's credit card market.

Years of experience in the development of foreign credit card markets have proven that the credit scorecard technology Credit Scorecard is a powerful means to curb credit card credit risks. A credit scorecard is actually a mathematical model used for personal credit risk control. It uses data mining technology to analyze a large amount of customer historical data accumulated by the card issuer, find out the characteristic values ??and patterns of customer credit risks, and establish corresponding mathematical models to assess risks for new credit applicants or existing customers. There are three types of credit scorecard models: application credit scorecard, behavioral credit scorecard and collection credit scorecard, which respectively provide pre-event, during-event and ex-post credit risk control for credit card business.

l Application Scorecard (Application Scorecard): The application scorecard model is specially used to evaluate the credit of new application customers. It can effectively and quickly identify customers through the relevant identity information filled in by the applicant. and classify good/bad customers, helping card issuers establish the first line - an ex-ante credit risk firewall.

l Behavior Scorecard (Behavior Scorecard): The behavioral scorecard model monitors and predicts the behavior of cardholders to achieve the purpose of assessing the credit risk of cardholders. Behavioral scorecard models can be used to automatically monitor and adjust credit limits, authorize and replace cards upon expiration, and predict bad debts. For example, if a customer wants to increase their credit card limit, the customer's previous spending and credit patterns need to be analyzed for approval using a behavioral scorecard model. Similarly, this scorecard model could be extended to the bank’s other personal credit products.

l Collection Scorecard (Collection Scorecard): The collection scorecard model is a supplement to the application scorecard and behavioral scorecard, especially established when cardholders have overdue loans or bad debts. Collection scorecards are used to predict and evaluate the effectiveness of actions taken on a particular bad debt, such as the likelihood that a customer will respond to a warning letter. In this way, the card issuing bank can take different effective measures to deal with overdue loans in different situations based on the predictions of the model.

As can be seen from the above, the credit scorecard model provides an objective and accurate assessment and control mechanism for the control of bank credit risks, especially personal credit risks. Data mining technology based on data statistical analysis can summarize and summarize the background characteristics of "good customers" and "bad customers" by collecting and analyzing a large number of customers' behavior, credit and background records according to the bank's business policies, including: Different attributes such as age, income, gender, housing conditions, marital status, occupation, education status, etc. can accurately calculate the spending power and repayment probability of customer groups with different attribute values, thereby establishing a system that can effectively distinguish between good and bad Mathematical models of customers help credit card institutions establish the first line of defense for credit risk prevention. For the bank's existing customers, the behavioral scorecard model can track and monitor each customer's behavior, consumption and repayment data, and intelligently adjust the customer's credit limit based on the corresponding model to deal with possible arrears, Bankruptcy can also be warned in advance. For bad debts that have already occurred, the model established by data mining technology can analyze and calculate the collection cost and recovery probability of the bad debt based on the customer's background and credit record, helping the bank to take correct and effective measures. It can be seen that the use of data mining technology can enable banks to effectively establish a credit risk control system before, during and after the event.

When establishing a credit scorecard model, the data used is mainly the data on credit card operations in the past period (1-2 years). The data includes two parts. One part is the background information of the cardholder, such as based on the answers on the application form: age, gender, marital status, educational background, characteristics of family members, housing situation, occupation, professional title, income status, etc.; the other part of the data is the cardholder’s past records. Information is displayed when using a credit card, such as its frequency of use, amount, repayment status, etc. In order to ensure the fairness and unbiasedness of credit card establishment, some information of rejected persons is sometimes needed to modify the credit scoring model.

Main mathematical analysis methods used to establish credit scorecard models

l CHAID Analysis / Tree Analysis

l Discriminant Analysis

l Logistic Regression (Conditional or Unconditional)

l Neural Network

2. Customer Segmetation

Faced with the intensifying competition in the credit card market and the gradual change of consumers To mature, banks must continue to provide consumers with effective personalized products and services in order to attract customers (especially high-end customers) for a long time, rapidly expand market scale, and truly create benefits.

Banks have begun to classify and analyze their customers a long time ago, but these classifications are mainly based on customer background information, including: classification according to income, age, region, etc. , can only statically and roughly reflect some background conditions and distribution of customers, and cannot allow card issuers to grasp the actual consumption behavior and preferences of customers. For this reason, in the "one-to-one" marketing market that advocates personalized products, customer segmentation supported by data mining technology has become an important marketing analysis method