Arabic text classification using principal component analysis with different supervised classifiers

Text classification (TC) is the process of automatically labeling a text based on its topic. Generally, TC suffers from data high dimensionality of the feature space. In this paper, we use the principal components analysis (PCA) as a feature extraction method for reducing the space features size with an application to Arabic texts. To evaluate the effect of this technique on the classification process, we use five popular classifiers: logistic regression, k-nearest neighbors, decision trees, random forest and support vector machines. The results show that the proposed method yields a substantial improvement of classification accuracy of most of the classifiers with reduction in execution time. To the best of our knowledge, it is the first time PCA is used for Arabic TC. In addition, the gain in training time is comprised between 5 and 800. Accuracy is improved in 60% of the cases.