Early Flood Risk Assessment using Machine Learning: A Comparative study of SVM, Q-SVM, K-NN and LDA

Abstract

Abundant floods and cyclones are the major cause of large emergency and acute ruin of properties in various countries. Usually floods are acknowledged as one of the most crucial problem in Malaysia, Indonesia, Bangladesh and France etc. Diverse techniques were carried out for a robust prediction system to investigate the flash floods. A dynamic system for the identification of run offs involves the computation of water peak, rainfall velocity, Global Positioning System-Precipitable Water Vapor (GPS PWV), wind speed, orientation, complex levels of river, land humidity, oceanic basement pressure and flash flood color with authentic cognizance algorithms. Accurate and precise forecasting of floods is very complex as it depends on many factors like precipitation, cloud to ground flashes, geomagnetic field, color of water, wind velocity, wind direction, temperature and others. In this research paper classification approaches like Linear Support vector machine, Quadratic Support vector machine, K-nearest neighbor and Linear discriminant analysis have been implemented to classify the true positive event of flash floods accurately and precisely. Comparative analysis has been performed between these three algorithms to determine the highest accuracy algorithm. Parametric comparison and results of training and testing proved that Support Vector Machine (SVM) performed very well.