Research On Naive Bayesian Classifier Algorithm

Classification is a very important research topic in data mining, it aims to constructa classification model or a classification function, which map the data of the data set toa given category. One of the hot topics of the current classification algorithm is thenaive bayes classifier. The reason lies in it has solid theoretical foundation,comprehensive data sample information and prior probability information. But it alsoexist fatal weakness, its attributes independence assumption is often violate the actualsituation. This will largely influence the classification performance.This paper through to introduce several kind of Bayesian classification model andthen analyses their characteristics, in order to make full use of the classificationadvantage of naive Bayes classifier model, according to the assumptions limitation ofstrict class conditional independence, the paper puts forward some improvement andobtained better classification effect. This paper introduced two different metrics tomeasure the degree of association among attribute, and introduced a new groupingtechnology which gather attribute to groups when attributes had correlation, in order toweake the attribute independence condition as groups of properties betweenindependence, this largely extends the Naive Bayesian classification algorithm and itspractical range.In order to prove that the proposed algorithm has better classification performance,Data simulation results are explained. at each end portion of the algorithm.