A Naive Bayes Classifier Based on Neighborhood Granulation

The naive Bayes is a classifier based on probability and statistics theory, which is widely used in the field of text classification. But the assumption of independence between features affects its classification accuracy. To solve this problem, this paper studies the theory of granular computing and proposes a naive Bayes classifier based on neighborhood granulation. The neighborhood discriminant function is introduced to perform single-feature neighborhood granulation for all samples to form neighborhood granules and multiple characteristic granules in a sample form a neighborhood granular vector. The operation rules of granular vector, a prior probability, and the conditional probability of granular vector are defined, and then a naive Bayes classifier based on neighborhood granulation is proposed. Experiments on some UCI data sets, using different neighborhood parameters to compare with the classic naive Bayes classifier, the results show that the method can effectively improve the classification accuracy.