Comparative Study of Stemming Strategy for Hadith Text Classification

摘要

Arabic language is a language has a deep meaning. Text Classification has become trending in this era due to increasing data on the Internet. However, there is only few researchers only work on Arabic language due to the difficulties of Arabic language. In this study, the researcher using Hadith Arabic Text Dataset with different Stemming strategy method. The researcher managed to collect 1253 documents of Hadith sahih from different kitab which are Bukhary, Muslim and Tirmizi. Convolutional Neural Neural with Linear Support Vector Machine were used as classifier in order to investigate the performance of the model. The experiment were conducted with three different Arabic strategy method; light stemming, Khoja Stemming and no stemming method. The non-stemming strategy managed to get higher accuracy compared to other strategy method which is 89.05% of accuracy.