Our topic is "Predicting Floods as a consequence of Deforestation."
Prediction of floods is a difficult task due to the erratic and unpredictable behaviour of nature and the dependence of flooding on geographical factors which are difficult to quantify to be used in traditional machine learning models. For tried two approaches to try and predict floods.
- A purely weather based prediction models using SVM, Random Forest and Neural Network.
- A GIS based model which takes various geographical factors into consideration to give the inundation depth of water for given river discharge using the HAND model(D.Tarboton et al. 2017)
Weather based predictions gave accuracies of upto 90% but showed signs of overfitting. GIS based results show promising results but are yet to be evaluated for accuracy.
