Flood Risk Prediction Using Reservoir and Rainfall Data Based on Machine Learning Techniques

In the recent years, the use of machine learning approaches has shown promising results in enhancing the processes of environmental risk assessment and disaster prediction models. The main of this aim of this research study is to develop an automated flood risk prediction model with the help of historical rainfall data and the level of the reservoir storage. The proposed system will use machine learning models such as Logistic Regression, Decision Trees, and Random Forest to predict the flood risk levels based on hydrological and meteorological parameters. A formatted dataset containing the historical rainfall and water level of the reservoir data was employed to train and test the prediction models, and data preprocessing and feature extraction methods were employed to improve the accuracy of the prediction models. The experimental outcome reveals that tree-based ensemble models are more accurate in predicting the flood risk levels compared to the existing statistical models, especially when dealing with imbalanced datasets. The proposed system validates the use of machine learning analytics in disaster management by developing an automated decision support system for the early detection of flood-prone situations.