A Novel Relevance Vector Machine Classifier with Cuckoo Search Optimization for Spatial Prediction of Landslides

Abstract

In mountainous regions, landslides are the typical disasters that have brought about significant losses of human life and property. Therefore, the capability of making accurate landslide assessments is very useful for government agencies to develop land-use planning and mitigation measures. The research objective of this paper is to investigate a novel methodology for spatial prediction of landslides on the basis of the relevance vector machine classifier (RVMC) and the cuckoo search optimization (CSO). The RVMC is used to generalize the classification boundary that separates the input vectors of landslide conditioning factors into two classes: landslide and nonlandslide. Furthermore, the new approach employs the CSO to fine-tune the basis function’s width used in the RVMC. A geographic information system (GIS) database has been established to construct the prediction model. Experimental results point out that the new method is a promising alternative for spatial prediction of landslides.