Result: Flood prediction using machine learning algorithms.

Flooding is a natural disaster that brings about huge material loss and death. Flash floods are distinguished by sudden heavy downpours usually in urban areas with poor drainage systems. Coastal areas and those near dams are the most affected. Flooding gets worse with poor planning and late forecasts but good forecasting reduces the impacts of floods. The study seeks flood prediction improvement through solution-based, data-driven approach employing machine learning algorithms. For our study we have used a thorough dataset from Kaggle that comprises of meteorological and hydrological data sets. This was done by balancing the dataset to take care of class imbalance as well as feature selection for relevant predictors. Out of these six models were implemented: Multiple linear regression, Light Gradient Boosting Method, Extreme Gradient Boosting Method, Decision Tree, Artificial Neural Networks, Random Forest and Support Vector Machine (SVM). Training and validation was done on each model using 80% training and 20% testing data respectively. The performance of each model has been measured in terms of accuracy, mean absolute error (MAE). Comparative analysis identified the most effective model for flood prediction, providing valuable insights into model accuracy and the critical role of localised data in anticipating and mitigating flood impacts. [ABSTRACT FROM AUTHOR]