Optimizing the Identification of Suitable Congregations for Preachers Using a GMM-PCA-BIC Hybrid Clustering Approach

The assignment of preachers to mosques in Riau Province faces significant challenges due to the geographic dispersion of mosques and inconsistent scheduling practices. Preachers often travel long distances, and mismatched assignments lead to inefficiencies and reduced effectiveness in delivering sermons. This study addresses these issues by leveraging advanced clustering techniques to optimize preacher assignments. Specifically, we applied Gaussian Mixture Models (GMM) to cluster mosque and preacher data based on geolocation, activity, and profiles. We integrated Principal Component Analysis (PCA) for dimensionality reduction to enhance the clustering performance, forming a GMM-PCA hybrid approach. Data was obtained from the Indonesian Ulema Council Riau Province Chapter, which consists of 185 mosques and 250 preachers, totaling 435 data points. The results indicate that using GMM alone, the optimal clustering was achieved with 10 clusters, yielding a Silhouette Score of 0.31. The GMM-PCABIC hybrid approach improved clustering quality. It achieved optimal results with 9 clusters, a significantly higher Silhouette Score of 0.48, and a lower BIC Score of 2696.18. These findings demonstrate the novelty of the GMM-PCA-BIC hybrid method in effectively capturing the complex characteristics of mosque and preacher data, ensuring efficient and relevant preacher assignments. Implementing this optimized clustering approach allows mosque administrators to make data-driven decisions, enhancing the effectiveness of religious activities.