Comparison between Deep Learning and Tree-Based Machine Learning Approaches for Landslide Susceptibility Mapping

The efficiency of deep learning and tree-based machine approaches has gained immense popularity in various fields. One model viz. convolution neural network (CNN), artificial (ANN) four models, namely, alternative decision tree (ADTree), classification regression (CART), functional logistic (LMT), were used for landslide susceptibility mapping the East Sikkim Himalaya region India, results compared. Landslide areas delimited mapped as inventory (LIM) after gathering information from historical records periodic field investigations. In LIM, 91 landslides plotted classified into training (64 landslides) testing (27 subsets randomly to train validate models. A total 21 conditioning factors (LCFs) considered inputs, each categorised under five classes. receiver operating characteristics curve statistical measures evaluate prioritise CNN achieved priority rank 1 with area 0.918 0.933 by using data, quantifying 23.02% 14.40% very high highly susceptible followed ANN, ADtree, CART, FTree LMT This research might be useful studies, especially locations comparable geophysical climatological characteristics, aid making land use planning.