PENGKLASTERAN DATA KUALITAS AIR TAMBAK MENGGUNAKAN METODE GAUSSIAN MIXTURE MODEL

This study aims to cluster pond water quality data to support decision making in fish farming management. The Gaussian Mixture Model (GMM) method is used as a probabilistic approach in clustering water quality parameters, namely pH, temperature, turbidity, and total dissolved solids (TDS). Data were collected from ponds in Kuala Kerto Village, North Aceh Regency, which is a traditional fish farming area. Before clustering, the data were cleaned from outliers and normalized using the Z-score method to improve the modeling quality. The model evaluation results showed that the GMM with 3 clusters provided the best results with a Silhouette Score of 0.55, Davies-Bouldin Index of 1.01, and the lowest BIC Score. Based on the standards for aquaculture water quality (pH 6.5–8.5, TDS <3000 mg/L, turbidity <300 NTU), each cluster was interpreted into good, moderate, and poor quality categories. Visualization of the results using PCA shows quite clear separation between clusters. This research provides a practical contribution in helping fish farmers monitor and evaluate pond water conditions in a more structured and data-driven manner. Keywords: Gaussian Mixture Model, Pond Water Quality, Clustering, Z-Score, Silhouette Score