A selective Bayesian classifier based on change of class relevance influence (CCRI SBC) was proposed by introducing a regulator factor into an attribute selection method

namely maximum relevance and minimum redundancy (mRMR).The regulator factor was used to change the influence degree of class relevance on the attribute selection

which can avoid the existence of redundant attributes in mRMR.In addition

a Bayesian information criterion was used to determine the optimal number of attributes automatically

which can overcome the randomness of classification results that easily caused by the setting number of attributes manually.In order to further make the CCRI SBC is applicable for continuous data

a discretization method

i.e.

equal frequency class attribute interdependent maximization was proposed

which has advantages of high classification correct rate and short discretization time.Experimental results on UCI datasets show that the proposed method can deal with the classification problem for discrete or continuous and high-dimensional data effectively.