A Detailed Comparative Analysis Towards Longer Terms Traffic Loads Forecasting with Autoregressive Integrated Moving Average (ARIMA/SARIMA) Models to Improve Transportation Services

Traffic flow forecast will be a great help in life of big cities citizens, especially in a time where road system cannot handle rush hours loads without constraining citizen to wait in traffic line and the population is in growth. The dataset used was captured by loop detectors that evaluate speeds length, time etc. In this paper, ARIMA and SARIMA were the models built for the prediction of the number of cars consisting of the traffic load. Although this study has been conducted for Tirana city, the methodology and discussions should be relevant to any big city. First, we transformed the time series to stationary with log scale transformation. After that, we found the right parameters for our models. Then we compared the results of two models: ARIMA (which we built with auto-Arima) and SARIMA, where Arima had the best outcome for the given dataset. The results were very satisfactory and with the Arima model we can make accurate forecasts for at least 3 months, showing that not only short-term forecasts are possible but even longer-term traffic load forecasting might be viable.