Study On The Credit Evaluation And Intelligent Decision-making System Of Credit Card Fraud Detection

In the last decade as China's credit card industry develops rapidly, the volume of card-issuing has been blowing out. However, with the rapid expansion of credit market, the high-risk feature of credit card has also been exposed. Currently the risk management and control of credit card has been an important subject in financial field. The classical technology of credit card risk is primarily established on the stochastic theory, the existing models of personal credit scoring consist mainly of single statistical measurement models. These models and algorithms have two basic problems: first, they lack capacity to handle nonlinear, ill-structured financial risk problems; secondly, these models ignore the randomicity and dynamic of risk factors. Because the financial risk measures are complex non-struture desision-making problems which have fuzzy character, modern intelligence technology such as artificial neural network has drawn much attention from researchers in recent years.As their own characters and technical defects, the single intelligent algorithm was bog down in application fields. It is a natural solution to research into the hybrid of these technology base on their common and characters. This research attempts to bring fuzzy systems and memetic algorithms into the artificial neural networks, aiming to construct fuzzy associative memory and create a new fuzzy neural network by incorporate fuzzy logic systems into neural network, which improving the learning ability of neural network as well as flexibility and intuitiveness of the network. At the same time, the global search capability of memetic algorithms is use to reduce the possibility of local minimum in artificial neural network, which will enable neural networks to have the double wisdom of learning and evolving. And reduction algorithm of rough set is use to reduce the complexity of the network structure, which will decrease the training time and avoid the“overfitting”problem. Finally a model based on memetic fuzzy neural network for credit scoring is proposed.According to the current situation of our country credit card services, after analyzed the character of credit card fraud, the paper put forward the credit card fraud models based on expert rules and memetic algorithm. Ultimately, an intelligent decision-making system is constructed to control credit card fraud risk.The research not only has considerable theoretic merit and well future of market in practice, but also has outstanding technical innovation.The major contribution of this paper are:1. A systematical summary of the personal credit scoring theory and relevant methods is done. Based on an in-depth analysis the shortcomings and performance of the existing methods, a fundamental thought of using intelligence technology to improve personal credit scoring models is bring forward;2. A detailed understanding of the research and development trends of intelligent computing; after an accurate grasp of the advantages and disadvantages of the neural network, memetic algorithms and fuzzy logic systems; The hybrid of the three technologies are explored correspondingly.3. After exploring the possible combination mode of neural network, fuzzy logic and memetic algorithms, an integrated model is constructed base on the hybrid intelligent computing technology.4. After an in-depth analysis of the principles and process of memetic algorithms, the improved designs of the memetic algorithm for integrated model's train are tried by fully exploiting the latest research achievements.5. A personal credit scoring model is proposed after analyzed and compared of the classical personal credit scoring technology.6. After an in-depth analysis of the credit card fraud mechanism, a intelligent decision-making system for credit card fraud detection is proposed based on expert rules and memetic algorithm.The research achievements and innovations mainly include:1. A modified memetic algorithm was presented through improving the global and local search technology in this paper. The proposed algorithm enhances the global search capability and improves the speed of convergence.2. Incorporating fuzzy logic into the neural network; optimizing the neural network structure by using memetic algorithms, then an integrated intelligence model is constructed base on memetic fuzzy neural networks.3. According to the sample data of credit card users, an integrated intelligence models for personal credit scoring are proposed by using rough set reduction as the pre-conduction system, and providing the practical algorithm procedures.4. The analysis and examination result of simulated experiment is done to confirm the remarkable success of the newly established model.5. The credit card fraud detection models are proposed based on experts rule and memetic algorithms, after that an intelligent decision-making system for credit card fraud detection is presented.