| Credit card is a kind of new payment means and credit tool. It has become a product that numerous commercial banks to launch it. Business credit card business have a high income but also accompanied by a high-risk. Along with our country enters WTO, the rapid development of e-commerce, credit system and credit payment environmental to be established, more and more consumers choose to use credit card payment. Therefore, guarantees the consumer credit card safe payment especially important. Contrast the international advanced credit card transactions risk management system, our countries commercial banks risk management of credit card transactions is still there is a great gap. Under the market economy condition, with economic development, there are more and more credit cards fraudulent, mainly as follows: fraudulent use of other people's credit transaction and malicious overdrafts, forgery of credit card fraud, invalid credit card fraud. Credit card fraud seriously disrupted the normal financial system, cause great losses both bank and credit card's owner, and impact on the healthy development of credit card business. Traditional fraud detection method usually based on Bayesian algorithms, decision tree algorithms and neural network algorithm. These three Learning algorithms of the actual results will depend on our understanding of the data and algorithm model setting. To solve these two problems, often depend on the user's prior knowledge and experience, cause that excessive rely on the user"skill".This thesis detailed analysis the characteristics of the credit card transactions, show that the application of data mining technology can be effectively found its internal hidden potential irregular transaction patterns. Support Vector Machine (SVM) is a new research area of data mining. It can better deal with these issues. This thesis apply support vector machine algorithm to set up a support vector machines based fraudulent transactions detection model to detect high-risk transactions. Specifically, do the following work: the credit card business data conversion needs for data mining data format, and data cleansing, discrete, missing data processing, establishing support vector machine detection model; Comparison of results of tests performed on various models.Finally, set up a support vector machine based credit card risk detection system, evaluation the experiment results, and proposed future research focus. |
