Data mining for hadith classification

Abstract

The holy Qur`an and Hadith are the two fundamental resources of the legislation and law in Muslim community. Including the Islamic books, these resources can be used as the sole authoritative source of knowledge and wisdom. Besides, they stand out as the source of a large collection of analysis and interpretation texts, which could provide a gold standard for artificial intelligent (AI) knowledge extraction and knowledge representation experiments. Recently, there are increasing attentions to automate the Islamic resources Qur`an, Sunnah and tradition books, motivate researchers to look for mechanisms that can represent and discover the knowledge of these resources. In the present study, extracted Islamic knowledge representing the focal point of the research, three famous books in Hadith science framed the corpus of this study. The present study attempted to explore new approach to classify Hadith according to its validity degree (Sahih, Hasan, Da`eef and Maudoo`) using data mining techniques, the proposed Hadith classifier (HC) model was built through learning process and was represented by the tree structure modeling. Moreover, the attributes of the instances originally were obtained from the source books directly. Whilst some of these attributes which is not mentioned in these books were indicated as null values, or missing values. A novel mechanism was employed to handle these missing data. This mechanism was generated based on the investigation methods of the Isnad in Hadith science. Representing or extracting Islamic knowledge is very critical step because it may affect life of Muslim, therefore, the results of the research were compared with the resource books, concurrently with the point of view of the expert in Hadith science. Indeed, the extracted knowledge shed light on the differences between Al-Imam Al-Bukhari, Al-Termithi and Al-Albani methods in takhareej AL-Hadith. Furthermore, the findings of the research showed that the performance of the proposed HC had significant effect with the proposed missing data detector method (MDD), the correct classification rate (CCR) was sharply increased from (50.1502 %) before using MDD to (97.597%) after applying it . Furthermore, the favorable results of comparing the performance of HC against naïve bayes classifier indicated that the decision tree (DT) Modeling is a viable approach to classify Hadith due to the excel performance, ease of implementation, and ease of rules induction and results interpretation.