Improving Urban Heat Island Predictions Based on Support Vector Regression and Multi-Sensor Remote Sensing: A Case Study in Malang City

The Urban Heat Island (UHI) phenomenon causes significant temperature increases in urban areas, adversely affecting the environment and public health. This research develops a prediction model of land surface temperature in Malang City using Support Vector Regression (SVR) with remote sensing data from Landsat-8, Sentinel-2, and SRTM. A cloud masking process is applied to improve image quality, while features such as NDVI, NDBI, NDWI, NDMI, elevation, and LST are calculated and normalized. The test results show that the Radial Basis Function (RBF) kernel with hyperparameters C = 10, Epsilon = 0.1, and gamma = 1 provides the best performance, with R² of 0.887, MSE of 1.625, and MAPE of 2.71%. This study shows that SVR with RBF kernel and appropriate tuning parameters can improve prediction accuracy. These results provide a strong basis for the development of more effective prediction models in managing UHI in big cities .