Travel Time Prediction Using a LSTM Model Reflecting Traffic State Transition in Interurban Expressway

Key Points

• The study aims to predict travel time based on traffic volume and speed data using deep learning techniques.
• Developed a travel time prediction model combining traffic flow dynamics and LSTM deep learning.
• Validated the model over a 35km stretch of highway during traffic bottlenecks and peak holiday travel.
• Measured prediction accuracy using Mean Absolute Error (MAE).
• Achieved 14 minutes MAE for 3-hour future travel time predictions during congested conditions.
• Demonstrated effective modeling that integrates current traffic data with predictive analytics.
• Validated the ability of the model to maintain accuracy despite variations in traffic conditions.