There are many existing problems in Hadith studies trending in the study field. The issues are changeable from the digitalization of the Hadith data to an exact case study of estimation of narrators' chain for a particular Hadith. However, in this paper, we are not concentrating on the such learning of estimating, confirming or authenticating a Hadith. It focuses more on the data mining use to the Hadith dataset. We put on the Hadith dataset onto one of machine learning tools which is text classification. The Hadith dataset is put into experiment for Hadith textual classification. It concentrates on the thematic classification based on the themes and words occurrences from the Hadith text (matn). The Hadith textual classification does not trace on the hukm and position or class of Hadith. This research does not categorize the Hadith into hukm Sahih, Hasan, Dhaif, or Mawdhoo'. However, the Hadith thematic dataset of this study use only Hadith from Sahih Bukhari, where all Hadith in the Book is categorized as sahih by Imam Al-Bukhari. The classification for this thematic Hadith dataset is implemented using Rapidminer, a machine learning tool using Naïve Bayes and Support Vector Machine (SVM) methods. From the results, the different value of accuracy for both SVM and Naïve Bayes Algorithm was 2.4%. The Naïve Bayes Algorithm displayed better result comparing to SVM. We believe that the result could be better by improving the data, algorithms, algorithm tuning or ensemble methods for the future experiments.