Research Of Classification In Data Mining Based On Bayes Technology

Classifying based on Bayes Technology has got more and more interests in the field of data mining. This thesis makes a study of two Bayesian classifying models which are Semi-Naive Bayesian Classifier and Increasing Bayesian Classifier.Semi-Naive Bayesian Classifier extends the structure of Naive Bayesian Classifier in order to get rid of the limit of the assumption of independence between feature attributes of Naive Bayesian Classifier and improve the performance of classification. The key of model learning of Semi-Naive Bayesian Classifier is how to combine feature attributes effectively. Since most algorithms are not effective and not very meaningful in combining, this thesis proposes an algorithm based on a kind of Semi-Naive Bayesian Classifier which is measured by conditional mutual information(CMI-BSNBC). This thesis implements the CMI-BSNBC model and uses it to carry out series of comparing experiments on experimental data ,with plenty of resultant data been obtained .After synthetically analyzing the experimental result we make some conclusion which show the effectiveness of the model.The key of Increasing Bayesian Classifier is the policy of how to choose test samples. This thesis studies how to make full use of prior knowledge and transmit it. The new model is presented which is based on the 0-1 loss of classification and uses training set to verify the test samples, which assures that the test sample more compatible with the training set be chosen firstly.