Research of full Bayesian classifiers based on tridiagonal matrix
Leng Cui-pin · Jisuanji yingyong yanjiu · 2015
Naive Bayes classifiers with continuous attributes can not make full used of dependency information between attributes. And in their dependency extension,it is difficult to calculate the inverse and the determinant of the high-order covariance matrix. To solve the dilemma,combining the tridiagonal matrix and multivariate Gaussian function can help build the full Bayesian classifiers with continuous attributes. And introduced smoothing parameters to tridiagonal matrix,which can contribute to the classifiers optimization. Experimental results resulting from using UCI data show that the optimized classifiers with continuous attributes have good classification accuracy.