Applications of the Maximum Entropy Principle in Spam Email Classification

摘要

For the small initial samples and highly variance of practical spam emails, maximum entropy principle is applied into the classification of spam emails. Incremental learning method is also integrated to increase the performance and smoothing method is utilized to process the sparse property of signature word. The simulation experiments show the effectiveness of our email classification.