An Improved Naive Bayes Algorithn
Xiaoyin Zhang · 1999
Although Naive Bayes algorithm, which is based on the assumption that every attribute is independently given the class, is shown to be surprisingly accurate for some classification tasks, yet in some large databases, the accuracy of Naive Bayes doesn't scale up other complicated algorithms. An improved algorithm is thus proposed which join the attributes according to their degree of dependency. Experiments show that this new approach out performs Naive Bayes algorithm in many databases, especially in those areas where decision tree algorithm such as C4.5 performs well.