Prediction of Flood in Bangladesh Using Different Classifier Model

Abstract

Bangladesh is highly affected by climate change scenarios notably floods due to its location on the world map in the South Asian region. Besides due to monsoon rains and high upstream rainfall in several areas eventually turn into floods. Thus, early flood forecasting might save human lives as well as agriculture crops. In this paper, we have applied different machine learning classifier models (Decision tree, Naive bayes, k-NN and Random forest) with a view to predicting the occurrence of flood. RapidMiner tool has been used extensively to perform data preparation, decision tree model generation, cross-validation, model selection and optimization of the model parameters. It is seen that the decision tree model has performed well by achieving an accuracy of 94.23% which is further optimized to reach 94.68%. Feature Selection using ‘correlation matrix’ is also a good aspect of this work by which we have achieved a good accuracy.