The TF-IDF term weighting is one of the most successful methods in feature extraction. Moreover, the highest term frequency (TF) for each document is concerned for feature extraction stage because it introduces information on text documents. A good model that was extracted by using supervised machine learning techniques that was selected the relevant features and determined how can convert them for a learning machine technique. This paper was focused on the statistical modeling and term weighting. Finally, the experimental results show that statistical modeling and term weighting method improves the performance of feature extraction. Simulation results show the superiority of the proposed technique. In general, statistical modeling and term weighting using TF-IDF method present a good results with respect to the evaluation metrics.