A Binarization Approach to Email Categorization using Binary Decision Tree

Binarization approaches are found promising in performing the task of email categorization. Amongst the standard binarization approaches, the one combining the some-against-rest binarization method and round robin assembling method is discovered most effective. However, two drawbacks are worth noting, i.e., effectiveness of the some-against-rest binarization method and computational complexity in training. This paper presents an algorithm in finding the binary decision tree, i.e. the optimal some-against-rest binarization solution. In classification stage, the binary decision tree is combined with the elimination method to address the two problems. Experimental results show that the binary decision tree is more effective in email categorization and computationally less complex in training.