Combining Climate Models Hydrological Analysis, and Machine Learning for Real Time Flood Prediction and Disaster Mitigation

Authors- Suruthi.K, Assistant Professor D.R.krithika

Abstract--Flooding has become an increasingly critical issue across the globe, driven by climate change, unpredictable weather patterns, and the overexploitation of natural water systems. Predicting flood probalities accurately is vital for effective disaster management and prevention. This project focuses on flood probability prediction using regression-based machine learning models, incorporating climatic and hydrological data to forecast flooding risks. The proposed model leverages climate-related factors such as rainfall intensity, river height variations, land use patterns, and drainage systems to predict the like hood of floods in a given region. The methodology involves multiple steps, starting with data collection and preprocessing, where the dataset is cleaned by handling null values, removing duplicates, and dropping irrelevant columns. The dataset is divided into training and testing subsets using train-test split methods. Standard regression models, including Linear Regression(LR), are implemented, and techniques like K-fold cross-validation ensure robust model performance. Model selection is followed by hyperparameter tuning, which explores the of L1 (Lasso) and (L2 Ridge) regularization to improve model accuracy and prevent overfitting. Decision tree regressors are also evaluated to compare model performance and explore non-linear
relationships in the data. After developing the model, evaluation metrics such as accuracy, precision, and recall are employed to measure its effectiveness.

DOI: /10.61463/ijset.vol.13.issue2.453