Learning Of Bayesian Network Classifier And Its Application In Credit Scoring

In this era of information explosion, the scale of data continues to surge, data mining technology has become an important means to obtain useful data information.The Bayesian network because of its solid foundation of probability theory and graph theory, visual representation and quantitative description ability, make it become one of the most representative of the intelligent information processing model. At present, the Bayesian network has been successfully applied in various of fields such as industry, agriculture, biology, medical and military, and also it has significant social benefits and economic benefits. So, it it very vital for the further study of Bayesian networks. This paper focuses on the research of Bayesian network classifier learning, following are the major works:First, the paper introduces the overview of the data mining and Bayesian networks.The thesis introduces the common tasks, basic process, classification and classification problem of the data mining; then it describes the actuality, background and applications of bayesian network, and then summarizes the characteristics of many classification models. After that, the advantages of networks are discussed compared with other methods.Second, the paper introduces kinds of Bayesian network learning algorithm, and then analyze the Bayesian network learning algorithm and some problems. In the learning part of network’s parameter, it mainly introduces three learning algorithm,but we focus on the learning of the network structure, and put forward three single dimensional Bayesian network classifiers algorithm based on local learning, they are IPC-GBNC, LAS-GBNC and LAS-GBNC+, and then we extends the single dimensional Bayesian network classifier and propose a learning algorithm named DOS-GMBNC, a multidimensional Bayesian network classifier, and basing on experiments, it fully demonstrate the superiority of the algorithm.At last, the credit scoring has a vital role in personal credit card business of commercial bank. In this paper, the dataset credit is provided by a foreign bank, thenuse LAS-GBNC+ to structure the Bayesian network model of the credit data. The experimental results show that this model is better than the Logistic regression model, NB model and J48 model. Therefore, this model to the bank credit analyst has a certain value.