Research On Construction Of Bayesian Network And Its Applications

Data mining constitutes the research on the theories and methods of mass data analysis. It is enjoying widespread application in retailing, military, business intelligence and finance. As a kind of network representing uncertain relationships based on probabilities, bayesian network provides a model to integrate the causal relationships. In artificial intelligence, it is an important method to deal with uncertainties. In the mean time, it can also be taken as a significant theoretical tool in data mining. Currently, bayesian network is playing an important role in medical diagnosis, risk evaluation, decision making and statistics.This paper firstly introduces the definition, tasks, functions and methods in data mining. Then it brings the definition, application and general ways of construction of bayesian network. As a major part of bayesian network construction, the classical search algorithms and network quality evaluation standards are also introduced in detail. Furthermore, it introduces Weka, a platform for data mining, with a focus on the structures and major functions of the Java classes associated with bayesian network. Based on these theories and technologies, this paper introduces a new network construction method called Combination Method, which picks different parent set for each node in the network based on the networks constructed by different classical methods and applies the measure approach MDL to evaluate the quality of network. Finally, the remaining part of the paper gives an example about its application on fund data analysis. The related specification of problem, variable choosing, data collection and processing, and the actual implementation of Combination Algorithm are shown. The network constructed can represent the variables in fund data and it can be used as a reference for investment analysts in the fund investment field.