<![CDATA[The Indonesian Ulema Council (MUI) of Riau Province plays an important role in dakwah (Islamic preaching) development, yet its evaluation methods remain limited. Understanding congregant feedback is crucial, but manually analyzing thousands of comments is ineffective. This research aims to apply topic modeling to automatically identify the main themes within congregant opinions. The algorithm used is Latent Dirichlet Allocation (LDA), analyzing 2,581 comments collected from the MUI Riau Smart Evaluation System. The research phase involved text preprocessing, such as cleaning, case folding, tokenizing, stopword removal, and stemming to produce clean data. This data was then converted into a Bag-of-Words (BoW) representation as input for the LDA model. The optimal number of topics was determined through evaluation using Coherence Score and Perplexity. Experimental results show that a configuration with 16 topics provides the best balance between semantic coherence and model generalizability, with a Coherence Score of 0.5008 and a Perplexity of -7.7787. The identified topics reflect diverse aspects, including prayers, appreciation for preachers, respect, discussions on Islamic values, and spiritual reflections. The LDA method proved effective in extracting thematic patterns from congregant opinions, providing a foundation for developing a real-time evaluation system.]]>
Copyrights © 2025
