Abstract
Nowadays, a huge number of web documents are available on the Internet, which makes the retrieval process of a specific topic very difficult, where some irrelevant pages may be retrieved as well. The automatic classification of web documents and pages has an essential application in different domains such as medicine, health, science, and information technology. A large number of web pages classification methods have been proposed to improve the search capabilities, especially in English language. In addition, the current classification methods attempt to classify the English web pages, and at the same time to reduce the high dimensionality of features extracted from these web pages. Due to the lack of classification methods for other languages, this paper focuses on Arabic web pages classification according to its scarcity as well as the importance of the Arabic language. In particular, we propose an evolutionary model based on binary particle swarm optimization (BPSO) combined with random weight networks (RWNs) as an induction algorithm to reduce the high dimensionality of features in the Arabic web pages and to perform document classification automatically. The datasets used in this paper were collected from popular Arabic websites. We collected three different datasets relating to three different fields, namely Computer Science, Science, and Health. Further, Taguchi method is incorporated to locate the best parameters of the proposed algorithm. The experimental results showed that the proposed model gives better performance results for Arabic web pages classification. In addition, an analysis study was conducted to identify the most important features learned from the proposed model as well as the most important tags. The results showed that list tag has obtained the highest percentage, which reflect its effectiveness on the classification of Arabic web pages.