Classifying Muslim ideologies from Islamic websites using text analysis based on Naive Bayes and TF-IDF

• Main Authors: Sharey Moustafa, Salma Moustafa, Olowolayemo, Akeem
• Format: Article
• Language: English; English; English
• Published: International Islamic University Malaysia 2022
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://irep.iium.edu.my/101738/1/Submissions%20_%20International%20Journal%20on%20Perceptive%20and%20Cognitive%20Computing.pdf; http://irep.iium.edu.my/101738/2/Classifying%20Muslim%20Ideologies%20from%20Islamic%20Websites%20using%20Text%20Analysis%20Based%20on%20Naive%20Bayes%20and%20TF-IDF.pdf; http://irep.iium.edu.my/101738/3/%5BIJPCC%5D%20Submission%20Acknowledgement%20-%20akeem%40iium.edu.my.pdf; http://irep.iium.edu.my/101738/; https://journals.iium.edu.my/kict/index.php/IJPCC
• Institution: Universiti Islam Antarabangsa Malaysia

Reliable digital Islamic information is one of the challenges faced by innocent Islamic information seekers such as young Muslims, new Muslims as well as others who desire to find authentic information about Islam, Prophet Muhammad (saw), and Muslims, in general. Several deviant ideologies abound, and they also present their information using the internet, sometimes involving digital deception. In the digital era, misleading Islamic information may affect people’s beliefs, behaviours, and attitudes. Many websites are equally based on several schools of thought regarding Islamic practices which could be difficult for the new Muslims, and the young generation of Muslims to recognize what to follow among these different websites based on the information presented on the sites. Some other variants of practices are considered to be deviants by the mainstream Sunni scholars which may be misleading for innocent Islamic information seekers including non-Muslims. Consequently, the need to categorize different Islamic websites based on different schools and branches becomes imperative. This initial study focuses classification of Islamic websites utilising website categorization and text classification approach to their textual contents. The proposed technique classified 60 Islamic websites into two various categories Sunni and Shia using TF-IDF for features extraction while using Multinomial Naive Bayes for classification. In addition, extracting the keywords for each of the two categories assisted in the classification process. The results show that Multinomial Naive Bayes was easily implemented and predicted the categories of Islamic websites with an accuracy of 0.89, precision 1.0, recall 0.80, as well as an F1 score of 0.89. The keywords that differentiate Sunni websites from Shia's websites were extracted. It was found that the best keywords that can be used in search engines to identify Sunni websites are Islam and Muslim, while Shia and Imam are the most prominent keywords that can be used to identify Shia's websites.