Abstract

Flood is one of the most dangerous natural disasters on earth and it cannot be completely prevented, however, accurate prediction can help save life and property. Our work focuses on evaluating and comparing the accuracy and effectiveness of using Random Forest (RF) technique and Artificial Neural Network (ANN) technique to predict the magnitude of flood. Both techniques were used to train and test the models using a dataset containing key hy drological variables, such as rainfall and temperature. The performance of these models was assessed using key evaluation/performance metrics including; accuracy, precision, recall, F1-score, confusion matrices, and ROC curves. The model developed using RF technique outperformedtheANNmodel. TheRFmodelachievedanaccuracyof97.0%,whiletheANN model achieved a 90.67% accuracy. It was further discovered that the RF model had better precision, recall, and F1-scores for both low and high flood severity levels. It was concluded that the RF offers more reliable and precise flood magnitude predictions, making it a more suitable option for real-world flood risk assessment and management. We recommend that future studies could explore further optimization techniques and the integration of additional hydrological variables to enhance predictive performance

Statistics