Landslide susceptibility mapping and management in Western Serbia: an analysis of ANFIS- and SVM-based hybrid models

Landslide susceptibility mapping (LSM) is essential for land-use planning, as it helps to identify areas at risk of landslides and enables effective prevention measures be taken. Various statistical machine learning (ML) models are used in LSM, including SVM ANFIS, which have shown promising results. However, determining model performs better remains a key challenge. To address this issue, paper aims compare six hybrid constructed with two well-known powerful ML models, namely three meta-heuristic algorithms, Genetic Algorithm (GA), Differential Evolution (DE), Cultural (CA), LSM case study western Serbia. In the process building 359 landslide sites 14 determinants were used. The accuracy was evaluated using several indexes, Root Mean-Squared Error (RMSE), coefficient determination (R2), Area under Receiver Operating Characteristic Curve (AUROC). modeling results showed that SVM-GA has highest (AUROC = 0.78) predicting incidence, followed by ANFIS-GA 0.775), SVM-CA 0.773), ANFIS-DE 0.771), SVM-DE 0.76), ANFIS-CA 0.65) validation phase. Therefore, suggested SVM-based more accurate than ANFIS-based thus, modelers may use such applications. This provides valuable insights into identifying most appropriate can help mitigate risks associated ensure sustainable practices.