In this study, we focus on the detection of coal mining sites using hyperspectral imaging technology, specifically in the visible and near-infrared (VNIR) regions of the electromagnetic spectrum. Hyperspectral imaging, which captures information across multiple contiguous spectral bands, offers a detailed way to observe subtle differences in surface materials. For this research, a hyperspectral image consisting of 17 contiguous spectral bands in the VNIR region was employed, which allows for high spectral resolution. Our main goal was to identify the most effective combination of normalized differential spectral indices (NDSIs) for accurately detecting coal mining sites. NDSIs are mathematical combinations of different spectral bands that enhance the detectability of specific materials by minimizing the influence of other features. To accomplish this, we utilized a random forest classification model, a machine learning technique known for its robustness and effectiveness in handling large datasets with complex feature interactions. The random forest model was trained using a ground truth binary classification map, which labeled areas as either coal mining sites or non-coal mining sites. Ground truth data, typically obtained from field surveys or high-resolution satellite imagery, plays a crucial role in training the classification algorithm by providing accurate reference points for comparison. One of the key performance metrics we utilized in evaluating the effectiveness of different spectral band combinations was the mean accuracy decrease. This metric highlights which bands or indices contribute most significantly to the overall classification accuracy. By analyzing the mean accuracy decrease, we were able to identify the optimal spectral band combinations in the VNIR region that were most effective in detecting coal mining sites. The results of this study were compared with previously established methodologies from the literature, and statistical analyses were conducted to further validate the findings. These comparisons allowed us to assess the improvements and limitations of our approach in detecting coal mining sites using hyperspectral imaging.