A Comparative Study on Arabic Text Classification

This paper focuses on Automatic Arabic classifications. Arabic language is highly inflectional and derivational language which makes text mining a complex task. In classifying Arabic text, there are many published experimental results. Since these results came from different datasets, authors and evaluation metrics, we cannot compare the performance of the experimented classifiers. In this paper, we compared six well known classifiers, which are: Maximum entropy, Naïve Bayes, Decision Tree, Artificial Neural Networks, Support Vector Machine ,and k-Nearest Neighbor using the same data sets and the same experimental settings. The recall , precision and fmeasure for the classifiers are computed and compared. Then, the comparison has been done after applying feature selection on Arabic datasest.