Landslide Risk Analysis Using Machine Learning Principles: A Case Study of Bukit Antrabangsa Landslide Incidence

Abstract

A quantitative analysis of landslides was carried out in this research to ascertain the risk that will be incurred on housing development built along hillside slopes. The study uses results from the most recent landslides analysis of GIS data and soft computing techniques to describe a concept for determining risk in a housing development situated on the hillside. The concepts presented by this study are timely, which means that computations define the time in which the housing development becomes vulnerable to landslides. The general idea enclosed in this research was to merge recent landslides analysis with real-life situations. The study location for this research was an already developed area situated along the hillside slopes of Bukit Antrabangsa, Malaysia. The approach includes predicting landslides using a novel machine-learning algorithm ensemble, the random subspace technique. The prediction process relates to the probability of occurrence of the slides responsible for the vulnerability of the housing development. A landslides prediction technique that employs GIS data to develop a geospatial database and machine learning algorithms for predicting future occurrences was used to produce the landslides inventory. The study area’s landslide inventory was then used as reference locations to predict landslide occurrence under the influence of ten landslide predisposing factors. Prediction results were evaluated for accuracy using the ROC (receiver operator characteristics) and calculated the AUC (area under the curve). Other parameters used to decide the quality of the models include the RSME (root-mean-square error), the MAE (mean absolute error), and the F-measure. The results obtained from the statistical analysis of the model show that the model has high predictive and success rates. The soft computing results now gave ways to obtain a timely probability of occurrence of the slides using a deterministic approach. The analysis was concluded by providing a conceptual equation to determine the timely probability of the slides using the available hazard information.