Research On The Applications Of Data Mining In Credit Card Fraud Detection

Recently with the economic growth and financial market opening-up around the world, governments have actively promoted various measures on the liberalization and internationalization of finance, resulting in that credit card market continues to grow up in both domestic and foreign countries. Transaction behaviors with credit card as the medium have increased and exceeded those with cash and bill as the medium. Credit card business has been an important part of banks income.However, credit card fraud transactions have gone up at a stupendous speed, and fraud methods have been retrodden and commit skills have been more and shrewder in the world, along with card circulation rising, economic scale expanding and market growing rapidly. Financial enterprises cannot detect fraud transactions effectively and bring increasingly loss, especially in China where social credit system has not been set up and exerted. How to detect credit card fraud transactions effectively, rapidly and exactly has been a commonly concerned problem in the financial industry at present.Firstly, this thesis introduces some knowledge of credit card and its business and risk management, deeply analyzes the causes of credit card fraud risk and the strategy of fraud detection and prevention. It also points out that research and discussion can be based on transaction data with transactions as a clue and on both client data and transaction data with customers as a clue.Then it studies classification models of data mining, focusing on two algorithms—support vector machine and decision tree, and also summarizes the literature of research on credit card fraud detection.Furthermore, aimed at two ubiquitous problems in classification study—skewed distribution and low-efficient single classifier, it puts forward k-means clustering for skewed distribution, combined classifier based on support vector machine and decision tree, adacost algorithm for classifier effect amalgamation.At last, using credit card transaction data of a commercial bank, it analyzes and detects fraud transactions based on the former models and compares the result with models without using k-means clustering and combined classifier. It proves that model built in this study is fit for credit card fraud detection.