Flood Prediction using Deep Spiking Neural Network

Abstract

The aim of this article is to analyse the Deep Spiking Neural Network (DSNN) performance in flood prediction. The DSNN model has been trained and evaluated with 30 years of data obtained from the Drainage and Irrigation (DID) department of Sarawak from 1989 to 2019. The model's effectiveness is measured and examined based on accuracy (ACC), RMSE, Sensitivity (SEN), specificity (SPE), Positive Predictive Value (PPV), NPV and the Average Site Performance (ASP). Furthermore, the proposed model's performance was compared with other classifiers that are commonly used in flood prediction to evaluate the viability and capability of the proposed flood prediction method. The results indicate that a DSNN model of greater ACC (98.10%), RMSE (0.065%), SEN (93.50%), SPE (79.0%), PPV (88.10%), and ASP (89.60 %) is predictable. The findings were fair and efficient and outperformed the other BP, MLP, SARIMA, and SVM classification models.