Semi-Supervised Support Vector Machine for Natural Hazards Forecasting. Case Study: Snow Avalanches

This paper explores the use of Support Vector Machine (SVM) as a predictive engine for natural hazards forecasting. It particularly discusses the issues of incorporating this classification method into a decision-support system for operational use in avalanche forecasting. The recent developments concerned with semi-supervised and transductive SVM-based learning targeted at applications in natural hazards forecasting on geomanifolds are presented. The real case study on spatio-temporal avalanche forecasting deals with the development of a predictive engine for the decision support system used at the avalancheprone site of Ben Nevis, Lochaber region in Scotland.