Ascertaining the impact of balancing the flood dataset on the performance of classification-based flood forecasting models for the river basins of Odisha

The climate shift being observed due to global warming has led to an increase in the frequency of natural hazards. Floods, which are the most recurrent and devastating of natural hazards, continue to take their toll on human lives and livelihoods. These losses could be avoided by designing models that can forecast floods at early stages, i.e., before they turn into disasters. This paper focuses on the designing of classification based flood forecasting models for the flood affected districts in the river basins of Odisha. Existing classification based models forecast floods using an imbalanced dataset. This paper attempts to ascertain whether balancing the flood dataset would result in the improvement of the existing classification based flood forecasting models. Experimental results showed that balancing the flood dataset using SMOTE and its variants have resulted in an improvement in the performance of classification based flood forecasting models.