Attribute weighted Naive Bayes classification based on a Space Search Optimization algorithm

Abstract:
Naive Bayes (NB) is an effective classification method. Due to its good performance, it is widely used to graphics and text classification problems in real-world applications. In this study, we propose an efficient improved model called space search optimization algorithm attribute weighted naive Bayes (SSOA-WNB), which combines attribute weighting method and a space search optimization algorithm (SSOA) method. In SSOA-WNB, the attribute weight is added to the naive Bayes classification formula, and the posterior probability is estimated by the attribute weighting method. To learn the attribute weight, we single out the SSOA method to estimate the weight matrix of the attribute value. We conducted a series of experiments on UCI benchmark data sets. The experimental results show that compared with the traditional NB method and some of the latest advanced algorithms, SSOA-WNB is significantly better than the compared well-known methods in terms of classification accuracy.