Extreme Event Forecasting Using Machine Learning Models

Extreme event forecasting helps in predicting the user demands during the peak travel times. The application of extreme event forecasting lies in predicting an increased demand for resources and hence can aid in effective resource allocation. The statistical approaches are used for the analysis of time series forecasting but for extreme events, it becomes difficult to predict the actual nature by using only the historical data. These methods alone are not sufficient to accurately predict user demands. Time series forecasting techniques along with machine learning algorithms are used to perform the extreme event forecasting. Here, in the paper, we have created the ensemble of two machine learning models, viz. recurrent neural networks (RNN) and Bayesian neural networks which remove the anomaly and improve the accuracy. Automatic feature extraction module long short-term memory (LSTM) has been used to extract the features. The proposed model enhances the accuracy by an extensive margin.