Publication

Abstract : Floods cause severe socioeconomic and environmental damage, making accurate predictions crucial for disaster preparedness. This study developed a machine learning based flood prediction model using the XGBoost algorithm, trained on district wise monthly rainfall data from 1970 to 2024, which was gathered from the Kerala Government Water Resource Information Portal. Kmeans clustering was applied to label the flood prone years based on historical rainfall patterns. Feature engineering techniques have been incorporated to improve predictive performance, including statistical aggregation, percentage contributions, moving averages, variability analyzes, and trend based transformations. Synthetic Minority Oversampling Technique was proposed to tackle the class imbalance. In Addition, a Long Short Term Memory model forecasts future rainfall, aiding proactive risk assessment. The proposed approach effectively captures nonlinear rainfall patterns, thereby improving the prediction accuracy. A comparative study using conventional approaches has confirmed its effectiveness. Metrics such as precision, recall, F1score, and accuracy are used to assess the effectiveness of the model. The results highlight the potential of integrating XGBoost, LSTM, and advanced preprocessing techniques for reliable flood prediction, supporting policy making and early warning systems, respectively.