Attribute weighting via differential evolution algorithm for attribute Weighted Naive Bayes (WNB)

The naive Bayes (NB) is a popular classification technique for data mining and machine learning, which is based on the attribute independence assumption. Researchers have proposed out many effective methods to improve the performance of NB by lowering its primary weakness-the assumption that attributes are independent given the class, such as backwards sequential elimination method, lazy elimination method and so on. Recently, Mark Hall presents a simple filter method for setting attribute weights for naive Bayes and proposes a decision tree-based attribute weighted method. In his paper, the experimental results show that the new weighted method performs better than other weighted methods. That weighting idea is taken as the objective of our study in which we use differential evolution algorithms to determine the weights of attributes and then use these weights in our previously developed Weighted Naïve Bayes (WNB). We evaluate the performance of new configuration (DE-WNB) on the whole 36 standard UCI data sets in Weka system. We also compare it with the decision tree-based attribute weighted methods and other methods mentioned in Mark Hall' paper for those data sets. Our experimental results show that the classification accuracy of our new algorithm DE-WNB is much higher than those of the other algorithms used to compare. The obtained classification accuracy is very good with respect to other common WNB classifiers in literature.