Kelantan daily water level prediction model using hybrid deep-learning algorithm for flood forecasting

• Main Author: Loh, Eng Chuen
• Format: Thesis
• Language: English; English; English
• Published: 2021
• Subjects: GB Physical geography
• Online Access: http://eprints.uthm.edu.my/6319/1/24p%20LOH%20ENG%20CHUEN.pdf; http://eprints.uthm.edu.my/6319/2/LOH%20ENG%20CHUEN%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/6319/3/LOH%20ENG%20CHUEN%20WATERMARK.pdf; http://eprints.uthm.edu.my/6319/

Flood, which is the most common natural disaster that occurs worldwide, causes massive casualties and damages to people and environment respectively. Hence, flood prediction is integral to minimise the damage and loss of life, while simultaneously aiding the government authorities and even the private sector in making accurate decisions when faced with incoming flood. Therefore, this present study had imputed the missing hydrological data using five imputation methods, namely Neural Network (NN), Moving Median (MM), Iterative Algorithm (IA), Nonlinear Iterative Partial Least Square (NIPALS), and Combined Correlation with Inversed Distance (CCID) imputation methods. Next, a newly developed hybrid deep learning (DL) algorithm is proposed to predict the daily water level in selected rivers that flow through Kelantan. The proposed model was then compared with two benchmark models, namely single Artificial Neural Network (ANN) and Wavelet Artificial Neural Network (WANN). The outcomes revealed that the MM imputation method resulted in higher accuracy with the lowest Root Mean Square Error (RMSE) for all rainfall and streamflow stations, in comparison to the other imputation methods. The experimental results portrayed that the proposed model achieved the best prediction accuracy in all performance measurements. The Mean Arctangent Absolute Percentage Error (MAAPE) results for all rivers ranged at 1-12%, which signified higher accuracy. Essentially, the proposed model may facilitate the government authorities and private sector to predict and plan better when dealing with the occurrence of flood.