5 hours flood prediction modeling using NNARX structure: case study Kuala Lumpur

Abstract

Flood is one of the most dangerous natural disasters that occurs frequently and can affect large community areas. It has becomes the main threat to people’s life and properties. Therefore, flood prediction has long been a popular subject matter to researchers around the world. This is because an accurate and reliable flood prediction is very much needed to provide early warning for residents nearby flood locations for evacuation purposes. Therefore, this paper proposed a 5 hours flood prediction modelling for Kuala Lumpur flood prone area using Neural Network Autoregressive Model with Exogenous Input (NNARX) and its Improved Modeling technique. The samples used for model training, model validation and model testing must be the data when flood events happened to obtain a good prediction model. All samples were real-time data requested from Department of Irrigation and Drainage Malaysia upon special request. Model validation and model testing were conducted to observe the prediction performance. The improvement technique was implement to improve the prediction performance from the original structure of NNARX model. The 5 hours NNARX flood water level prediction model and the Improved NNARX model have been successfully developed, analyzed and tested using MATLAB Neural Network Toolbox. Results show that the NNARX model successfully predicted flood water level 5 hours ahead of time.