Short-term traffic speed prediction under different data collection time intervals using a SARIMA-SDGM hybrid...

Area of Science:

• Intelligent Transportation Systems
• Traffic Engineering
• Data Science

Background:

• Short-term traffic speed prediction is crucial for proactive traffic control.
• Intelligent transportation systems (ITS) rely on accurate traffic data.
• Optimizing data collection intervals impacts prediction accuracy.

Purpose of the Study:

• To investigate short-term traffic speed prediction accuracy using varying data collection time intervals.
• To propose and evaluate a novel hybrid model for traffic speed prediction.
• To compare the proposed model against existing prediction methods.

Main Methods:

• A hybrid Seasonal Autoregressive Integrated Moving Average plus Seasonal Discrete Grey Model (SARIMA-SDGM) was developed.
• Traffic speed data was collected from an urban freeway in Edmonton, Canada.
• Model performance was benchmarked against SARIMA, SDGM, ANN, and SVR models.

Main Results:

• The SARIMA-SDGM model demonstrated superior performance, achieving the lowest MAE, MAPE, and RMSE.
• Prediction accuracy improved as the data collection time interval increased.
• Stable prediction accuracy was observed for time intervals exceeding 10 minutes.

Conclusions:

• The SARIMA-SDGM model is highly effective for short-term traffic speed prediction.
• Increasing data collection intervals enhances prediction accuracy in ITS.
• The findings provide valuable insights for optimizing traffic data collection strategies.