Clustering Analysis of Temperature Humidity Index (THI) andSupporting Weather for Food Crop Cultivation in Tegal

Abstract

The food crop agriculture sector faces serious challenges due to global climate change that disrupts the stability of conventional production systems. The Temperature Humidity Index (THI) is a crucial indicator for measuring environmental thermal comfort, which, combined with supporting weather parameters, can map the risk of crop failure. This study aims to analyze the clustering of THI and weather variables in the Tegal area using a Machine Learning approach. The dataset used is daily historical weather data for 10 years (2016–2025) from BMKG, including temperature (T), relative humidity (RH), solar radiation (SR), wind speed (WS), and rainfall (RF). The method includes preprocessing, normalization, THI calculation, and clustering using K-Means and DBSCAN. K-Means identified agro-climate vulnerability zones: Optimal, Alert, and Critical for food crop growth. DBSCAN effectively detected dominant cluster patterns and outliers of extreme weather anomalies. Internal evaluation shows K-Means performs better, with a Silhouette score of 0.4057 and Davies-Bouldin Index of 0.8391, compared to DBSCAN with 0.3957 and 2.7432. The results are expected to support farmers and policymakers in determining adaptive cropping patterns and mitigating climate change impacts in Tegal.