Research And Application Of Naive Bayesian Classifier

Naive Bayesian classifier is a simple and efficient classification algorithm, which is widely used in data mining and pattern recognition, but the na?ve Bayesian assumption often does not hold in the real world, affecting the classification results more or less. A number of techniques have explored simple relaxations of the Na?ve Bayesian assumption in order to enhance accuracy, but usually lead to the computational cost greatly.The major tasks and innovation points of this paper are as follows.Study the naive Bayesian classifier and its various improved model in detail, explore how to better learning the naive Bayesian classifier, propose a Naive Bayesian Classifiers using feature weighting based on rough sets, which improves the naive Bayesian classification performance.Propose a Naive Bayesian classification algorithm using feature weighting based on rough sets (FWNB). The feature weighting coefficients are directly induced by rough lower approximation of attributes, and can be regarded as the significance of each attribute when evaluating the posterior probability of the particular class value. Compared the FWNB algorithm with the naive Naive Bayesian classifier, Bayesian Networks and NBTree, and with the weighted Naive Bayesian Classification Algorithm Based on Routh Set in literature 26, a na?ve Bayesian classifier algorithm based on the rough set in literature 32, experimental results show that in most data-sets, FWNB algorithm has a higher classification accuracy at the cost of less computation.The FWNB classifier is designed to the guidance of Germany credit in order to validate its classification accuracy. Compared with the neural network model by West, and Xu-sheng Li et al. proposed an extension of the Tree Augmented Naive Bayesian network, Experimental results show that, FWNB is significantly better than other algorithms in the classification accuracy.