Services on Demand
Journal
Article
Indicators
- Cited by SciELO
- Access statistics
Related links
- Similars in SciELO
Share
Revista Tecnología en Marcha
On-line version ISSN 0379-3982Print version ISSN 0379-3982
Abstract
RIVERA-PICADO, Cristal and MENESES-GUZMAN, Marcela. Vehicle traffic flow forecasting Costa Rica highway 27. Tecnología en Marcha [online]. 2022, vol.35, n.4, pp.138-148. ISSN 0379-3982. http://dx.doi.org/10.18845/tm.v35i4.5892.
Forecasting vehicle traffic flow is considered an important input for traffic planning and management for the countries' intelligent transport systems (ITS). This article analyzes the hourly flow of light vehicle traffic that drives in highway 27 of Costa Rica in one direction (San JoseCaldera). The data collected by the ITS of the route is used to forecast the behavior of hourly vehicular traffic. For this, three forecasting methods are proposed, which are compared to select the model with best performance: Seasonal Arima (SARIMA), Seasonal Naïve (SNAIVE), and Autoregression with Neural Network (NNAR). All three models are evaluated and are considered useful for prediction, however the NNAR model results in better performance when forecasting the hourly time series with the lowest MAPE of 9.4 and is consider a candidate for use in ITS. By applying the cross-validation process in the models, the conclusion is supported that as the NNAR is tested for more days, the prediction results are more stable and accurate.
Keywords : Traffic flow forecasting; Seasonal ARIMA(SARIAM); Seasonal Naïve (SNAIVE); Autogression with Neural Networks (NNAR).