The Machine Learning to Detect Drought Risk in Central Java Using Landsat 8 OLI Remote Sensing Images

This study aims to develop a framework to detect the drought risk using the Machine Learning (ML) and vegetation index data extracted from LANDSAT 8 OLI images. Experiments were carried out through three stages: (1) data acquisition, (2) data preprocessing, and (3) analysis and modeling. The vegetation indices in these experiment are NDVI, TCI, VCI, VHI, and SAVI. The Meteorology data are Standardized Precipitation Index three month (SPI-3) data. The correlation analysis was made using Pearson Correlation and data classification was done using ML algorithms: K-NN, CART, RF, SVM and LDA. The results show that the vegetation index data formed a cluster pattern. The index vegetation had a positive correlation, which means that the higher the health of the vegetation canopy. This indices show the performance classification using ML methods for drought risk areas indicate that the ratio of valid vegetation indices with the total amount of data analyzed varies between 98% - 100%. The Cohen's Kappa Statistic test is between 0.8 - 1, and the Logarithmic Loss performance is between 0.037 - 0, 17.