Modeling Traffic Vehicular Counts at an Urban Intersection Using Complex Seasonal Modeling Techniques

The accurate estimation of traffic counts is critical for Intelligent Transportation Systems (ITS). In this paper, we introduce novel model-based parametric techniques for modeling traffic counts with complex seasonality at urban intersections. The three distinct models to fit and predict vehicular counts with multiple seasonal effects are dynamic harmonic, BATS, and TBATS. We have used loop detectors to collect hourly data over 71 consecutive days. All three models effectively capture seasonality not only within a day, but also encompass hourly, daily, and weekly seasonal effects. In contrast, the traditional SARIMA model fails to capture more than one seasonal effect in the data. Furthermore, the decomposition of the BATS and TBATS components provides estimations for hourly and daily effects, enabling a deeper understanding of the temporal flow of vehicles and leading to more efficient resource allocation and accurate travel time predictions. Next, we assess the performance of these three approaches across 11 different prediction windows, ranging from hours to days. Our results indicate that changes in the prediction windows do not significantly affect the predictability of the BATS and TBATS models. The BATS model outperforms the others in terms of the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Specifically, the MAE ranges from 2.37 to 27.39, and the MAPE varies between 0.07 and 0.72 for the BATS models varied windows. These findings support the practical application of the BATS model in traffic control rooms by implementing real-time traffic signal controls.