Text classification is a very important task due to the huge amount of electronic documents. One of the problems of text classification is the high dimensionality of feature space. Researchers proposed many algorithms to select related features from text. These algorithms have been studied extensively for English text, while studies for Arabic are still limited. This study introduces an investigation on the performance of five widely used feature selection methods namely Chi-square, Correlation, GSS Coefficient, Information Gain and Relief F. In addition, this study also introduces an approach of combination of feature selection methods based on the average weight of the features. The experiments are conducted using Naïve Bayes and Support Vector Machine classifiers to classify a published Arabic corpus. The results show that the best results were obtained when using Information Gain method. The results also show that the combination of multiple feature selection methods outperforms the best results obtain by the individual methods.

The process of eliminating irrelevant, redundant and noisy features while trying to maintain less information loss is known as a feature selection problem. Given the vast amount of the textual data generated and shared on the internet such as news reports, articles, tweets and product reviews, the need for an effective text-feature selection method becomes increasingly important. Recently, stochastic optimization algorithms have been adopted to tackle this problem. However, the efficiency of these methods is decreased when tackling high-dimensional problems. This decrease could be attributed to premature convergence where the population diversity is not well maintained. As an innovative attempt, a cooperative Binary Bat Algorithm (BBACO) is proposed in this work to select the optimal text feature subset for classification purposes. The proposed BBACO uses a new mechanism to control the population’s diversity during the optimization process and to improve the performance of BBA-based text-feature selection method. This is achieved by dividing the dimension of the problem into several parts and optimizing each of them in a separate sub-population. To evaluate the generality and capability of the proposed method, three classifiers and two standard benchmark datasets in English, two in Malay and one in Arabic were used. The results show that the proposed method steadily improves the classification performance in comparison with other well-known feature selection methods. The improvement is obtained for all of the English, Malay and Arabic datasets which indicates the generality of the proposed method in terms of the dataset language.