Research On Directed Forest-Augmented Algorithm Based On Bayesian Network

With the widespread application of databases, the scale of databases expands dramatically, and data mining has become a focus of research. The algorithms of data mining include association analysis, classifying and forecasting, and clustering analysis. Classification is one of the major parts of data mining which has aroused general concern due to its extensive usage.Researchers become interested in classification because it's of great importance in data mining. Many new classification algorithms, especially algorithms based on Bayesian network, have been used to construct classifier.Firstly, this paper introduces basic concept, background, methods and classification techniques of data mining technology. Then the paper defines construction of Bayesian network, and illustrates the principles of classification using a Bayesian network model. On the basis of comprehensive study of several Bayesian network models, the paper analyzes and estimates the advantages and disadvantages of these classification models. Lastly, the paper presents two kinds of directed Forest-Augmented classification models--directed Forested-Augmented model based on nodes sorting and directed Forest-Augmented model based on edges selection. The effectiveness of classification may differ for data sets with different characteristics because the models have different focuses. Therefore, the data should be analyzed roughly before classification and proper classification model should be selected according to the characteristics of data sets. Experiments show that the algorithm in this paper is superior to others in some data sets, and not inferior to others in most data sets.