Assessment of Landslide Susceptibility Using the PCA and ANFIS with Various Metaheuristic Algorithms

Abstract

It is very important for the susceptibility assessment and disaster prediction of the region to effectively evaluate the landslide susceptibility. In this study, Particle Swarm Optimization (PSO), Artificial Bee Colony algorithm (ABC), Shuffled Frog Leaping Algorithm (SFLA) and Bat algorithm (BAT) are used to optimize Adaptive Neuro-Fuzzy Inference System (ANFIS) to evaluate the landslide susceptibility. 811 sample points were collected through remote sensing analysis and field investigation for susceptibility analysis. Fifteen landslide evaluation factors were quantified and normalized, and the Principal Component Analysis (PCA) method was used to compress them into 6 main factors. The accuracy analysis results of the area under the curve (AUC), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) evaluation models show that the AUC values of PSO, ABC, SFLA and BAT are 93.6%, 96.2%, 90.8% and 86.1%, respectively. Among them, the accuracy of ABC is the highest. This study effectively evaluates the landslide susceptibility through a new neural network hybrid method, which provides a theoretical basis for landslide disaster susceptibility management.