Flood Damage Analysis Using Machine Learning Techniques
Snehil, Ruchi Goel
Abstract
In Whole World Flood is considered as recurring disaster which affect life and Economy both at very large extents. India’s statistics of damage caused by flood is in billion dollars every year. Thus, flood risk management where risk is measured in terms of damage (value in cost) is very important to reduce this annual loss of life and property. Most of the research is based on water depth used as a single variable for flood damage modelling [1]. Just like Korea in India heavy rainfall is the major reason for floods. In this Paper using Machine Learning techniques we are implementing Heavy Rain Damage Prediction Model as in [2]. Regression analysis is performed using models like Gaussian Na¨ıve Bayes, tree-based approaches and K Nearest Neighbour, etc [8]. The analysis was run on three states Bihar, Uttar Pradesh and Kerala. Among all the selected supervised learning technique Random forest and KNN performed best. The results may vary for states where other parameters related to cyclones, dams etc are also the primary reasons for flood.