Water Level Prediction for Disaster Management Using Machine Learning Models

A flood is an overflow of water and becomes the common natural disaster. Prediction of a flood is one of the challenges for disaster management around the world especially in developing countries. Thus, more accurate flood prediction models have been investigated according to the geographical locations. In this paper, we have studied and compared some useful machine learning models such as KNN, SVR and Linear Regression for getting better water level prediction. The proposed approach is applied to Ayeyarwady river in Myanmar. The future water level is predicted based on the time series data of past water levels. By the experiment, KNN (K-Nearest Neighbour) model shows the least mean absolute error and the error rate is just 0.17%. The predicted output of the proposed model agrees in the actual water level. Therefore, KNN model can be the potential solution for successful water level forecasting application in Ayeyarwady river. Keywords— water level prediction, time series analysis, KNN, SVR, Linear regression