Hybrid ML and DL models for flood level prediction

Abstract

Flood occurrences happen in a particular place(that particular place may be anywhere) due to many causes, it may be due to dam’s outbreak, because of poor infrastructure design aid during heavy rainfall times, Massive Rainfalls in a particular area, snowmelt in northern areas, and climate changes. In this work, Heavy Rainfall is considered as the major factor or key attribute for predicting the floods, particularly, the rainfall downpour amount for each month and annual rainfall value to predict flood occurrence in a particular place and in that particular season. In the work, Artificial models namely KNN Classifier, Logistic Regression, Decision Tree Classifier, Random Forest Classifier and Ensemble Learning and Deep learning architectures i.e. LSTMs (Long-Short Term Memory) and MLP (Multi Layer Perceptron) for predicting floods by comparing the accuracy obtained by ML models and Deep Learning models. The analysis is going to be done in three parts, the first part is the comparison of testing and training accuracies between each and every Machine learning models i.e. 5 state-of-the-art models which are mentioned above, the second part is to compare the test and train accuracies between the two deep learning architecture which are mentioned above.