Comparative analysis of machine learning algorithms in water extraction

Abstract There are many traditional methods available for water body extraction based on remote sensing images, such as normalised difference index (NDWI), modified NDWI (MNDWI), and the multi-band spectrum method, but accuracy of these is limited. In recent years, machine learning algorithms have developed rapidly been applied widely. Using Landsat-8 models decision tree, logistic regression, a random forest, neural network, support vector method (SVM), Xgboost were adopted in present research within algorithms. Based this, through cross validation grid search parameters determined each model.Moreover, merits demerits several discussed comparative analysis was performed with three determining thresholds NDWI. The results show that network has excellent performances stable model, followed by SVM regression algorithm. Furthermore, ensemble including forest affected sample distribution model tree returned poorest performance.