Flood Prediction by using Artificial Neural Network: A Case Study in Temerloh, Pahang

Floods are natural disasters that can cause significant property damage and sometimes result in loss of life. In Malaysia, floods occur every year, particularly on the East Coast of Peninsular Malaysia, due to the Northeast Monsoon and the impacts of climate change, which can lead to heavy rainfall at the end of the year. Temerloh, a district in Pahang, frequently experiences flooding events, especially between November and January. Despite various efforts in flood mitigation and preparation, the damage to both citizens and property each year results in costs amounting to thousands of Ringgits and the time needed to clean up the aftermath of floods. To address this issue, this research examined the hydrological and meteorological factors contributing to the floods in Temerloh and developed a machine-learning model capable of predicting future flood occurrences. The study utilized a dataset from the National Hydrological Network Management System (SPRHiN), which includes hydrological data and meteorological information for the specific location. The correlation analysis revealed a strong relationship between stream flow and water level to floods, with correlation coefficients (r values) of 0.83 and 0.76, respectively. In contrast, temperature exhibited an inverse relationship with floods, showing a correlation value of -0.28; this suggests that lower temperatures are associated with a higher likelihood of rain and subsequent flooding. The results indicated that the model, developed using an artificial neural network (ANN), achieved an impressive accuracy of 0.9909 and demonstrated strong performance, as evidenced by an area under the receiver operating characteristic (ROC) curve (AUC) value of 0.888. The model also exhibited low error rates, with a mean squared error (MSE) of 0.009 and a root mean squared error (RMSE) of 0.096. Additionally, the R² value of 0.768 and the F1 score of 0.875 indicate that the model possesses high precision and recall. Furthermore, a flood monitoring dashboard was created to provide interactive data visualization. This research is essential for understanding the factors contributing to flooding in Pahang and will offer valuable insights for future studies on floods.