<![CDATA[Tiga Saudara Catfish Farm in Bengkulu City has experienced varying mortality rates across different cultivation periods; however, these patterns have not been systematically analyzed using a data-driven approach. This study aims to identify catfish mortality patterns using Agglomerative Hierarchical Clustering (AHC) method. The dataset consisted of 90 days of operational records, including water temperature, pH levels, and feed types. The analytical process followed the Knowledge Discovery in Databases (KDD) framework, encompassing data preprocessing through One-Hot Encoding and Min–Max normalization, Euclidean distance calculation, and the implementation of the AHC algorithm using Single, Average, and Complete Linkage methods. The results indicate that the Average Linkage method achieved the best performance, with a Cophenetic Correlation Coefficient (CCC) value of 0.9789. The optimal number of clusters was determined to be four (k=4) based on a Silhouette Score of 0.8175. These four clusters represent distinct operational conditions within the farming system, namely Safe, Alert, Critical, and Recovery conditions. The findings further reveal that the use of alternative feed derived from bread waste contributed to a decline in water pH levels to a range of 6.0–6.8 and a significant increase in fish mortality, resulting in a total loss of 1,465 fish and an estimated financial loss of IDR 1,044,000 per production cycle. Model validity testing through White-Box Testing produced a CCC value of 0.9789 and a Silhouette Score of 0.8175, both exceeding the predetermined acceptance thresholds. Furthermore, system functionality testing through Black-Box Testing on 14 test cases within the AHC pipeline and Power BI dashboard demonstrated a 100% success rate. The clustering results were visualized through an interactive Microsoft Power BI dashboard, providing risk maps, daily mortality trends, and cluster distributions as decision-support tools for catfish farmers. This enables rapid identification of risk conditions and facilitates preventive actions before mass mortality events occur]]>
Copyrights © 2026
