Land Use Classification with different Machine Learning technique with Landsat MSS Image

Land use/land cover changes had accelerated near the turn of the 20th century owing to rapid and unregulated population growth, economic activity, and industrialization, especially in emerging nations. A quantitative assessment evidently is imperative for grasping and controlling land transformation processes. In this study, an assessment is made of the performance of four commonly used machine-learning algorithms-Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN) and Spectral Angle Mapper (SAM), for LULC mapping. Kappa statistics used to assess the accuracy of the study helps to see that all classifiers present accurate results, though some variation is present. The highest accuracy (0.89 Kappa) is attained through the Random Forest algorithm. These results highlight SVM and RF out of all the algorithms used for LULC mapping as the most effective classifiers. However, more tests with the RF algorithm should be conducted in various regions and climate settings to confirm its robustness and potential applicability.