Urban Pluvial Flood Forecasting using Open Data with Machine Learning Techniques in Pattani Basin

Pluvial floods are rare and dangerous disaster with small duration, which have the most destructive impact within urban areas. This paper explores opportunities of machine learning methods for forecasting of flooding phenomena in Pattani river using open data. The study factors include data collection period and location and configuration of prediction models. The analysis of quality characteristics for several machine learning-based algorithms and the set of upstream and downstream flood models was tested for different values of forecast steps. As the result, the Bayesian linear model was proposed for Pattani flood prediction. It can be used for reconstruction of historical rivers floods and forecasting of potential extreme events.