Development of a Flood Hazard Prediction Model Using Artificial Neural Network

Abstract

Flooding has become a major problem in most cities, and the need of a system that maps flood hazard is extremely useful to assess potential consequences, mitigate impact of large flood events, disaster preparedness, post-flooding relief operations, and emergency response team planning. This work describes the development of a machine learning prediction model that can be used to predict the severity of flood within Dumaguete City. The model was created using historical atmospheric, meteorological, flood, ground elevation, topographic, and geological data. Two preliminary prediction models were created using feedforward neural network and self-organizing map to map areas vulnerable to flooding within Dumaguete City. Fifty (50) different scenarios of selected dates from 2000 to 2014, the years where extreme flooding was experienced in the city, were tested to the two predictive models to reproduce flooded areas for verification. Observation on the tests performed on both models results to 80% and 86% prediction rate, respectively.