Traffic Flow Prediction Using LSTM, ARIMA & SARIMA Model for Intelligent Transportation System

Accurate and timely prediction of traffic data is an essential component in the planning, administration, and control of urban traffic in Intelligent Transportation Systems (ITS). Our study is dedicated to exploring three distinct models: LSTM (Long Short Term Memory), ARIMA (Autoregressive Integrated Moving Average) with parameters ($p, d, q$) that specify autoregression, differencing, and moving average orders, and SARIMA (Seasonal Autoregressive Integrated Moving Average), intended for analyzing seasonal time series data. ARIMA and SARIMA models are conventional statistical techniques employed for the study of time series data. They provide reliable methodologies for forecasting future traffic patterns through the analysis of historical data. To choose the most efficient model among the three options, we utilize performance measurement procedures to evaluate their predictive abilities. We utilize an external dataset obtained from the U.S. Government Open Repository. This dataset is classified as geographical data. Upon analysis, it was shown that all three SARIMA models had superior performance.