Bayesian Network Based On Gene Expression Programming And Its Application In Data Mining

Data Mining (DM) is the process which extracts the hidden, unknown, implicit information and knowledge form the large, incomplete, noised, fuzzy, random data. With the developing of communication and the computer technology, the data explosion become severely, in fact, we are drowning in data but starving for knowledge, which pulls the demands of powerful data analysis tools. The emergence of data mining provides strong technical support for the urgent need.More and more people convince that Data Mining can transform the raw data to useful pattern, which can extract huge commercial profit and scientific knowledge. Its application has spread from business to medical, military, commercial fields. After the development of more than a decade, DM has set up a solid foundation based on association rules mining, classification rules mining and cluster rules mining. It has combined the database, statistics, AI, visualization, and IT.Classifier is finding a concept description form the training data, which represented the data's integer information, and form the classifier we can predict the unlabeled instance. This paper presents a new hybrid model which combines the Evolutional algorithm and Bayesian Network, and applies the hybrid pattern to classification.Evolutionary Computation (EA) is kind of efficient, parallel, global, random searching method which use the reference of follow the rule of survival of the fittest, it acquire well performance in many domain. There are three branches: (1) Genetic Algorithm (GA) is the important role in EA, (2) Genetic Programming (GP) is a variant method of GA, (3) Gene Expression Programming (GEP) is a new genetic algorithm which invited by C.Ferreira. GEP combines the merit of GAs and GEP, and overcome their drawback. It uses the especial coding and transferring system, which artfully combine the genotype and phenotype. GEP has better performance than GP in symbolic regression, sequence induction, time series etc.Naive Bayes (NB) is a quickly, efficient method for classifier, but its attribute independent is always violated in real world. Augmented Bayesian Network (ANB) consider the attribute dependence, but the search space of Bayesian network is become abnormal huge as the number...