A Novel Segment-Based Approach for Improving Classification Performance of Transport Mode Detection.

Area of Science:

• Urban planning and transportation science
• Mobile sensing and machine learning
• Human mobility pattern analysis

Background:

• Effective transportation planning requires understanding urban mobility patterns.
• Smartphone data offers a valuable resource for monitoring daily travel activities.
• Existing transport mode detection methods often suffer from misclassification issues.

Purpose of the Study:

• To introduce a novel segment-based transport mode detection architecture.
• To improve the accuracy of traditional machine learning classification algorithms for transport modes.
• To develop a post-processing algorithm, the Healing algorithm, to correct misclassifications.

Main Methods:

• A segment-based transport mode detection architecture was designed.
• A post-processing algorithm named the Healing algorithm was developed to refine classification results.
• A mobile application was implemented to test the proposed architecture and algorithm.

Main Results:

• The Healing algorithm demonstrated up to a 40% improvement in classification accuracy.
• The implemented mobile application achieved a 95% success rate in predicting eight transport modes.
• The multi-tier architecture and Healing algorithm significantly enhanced transport mode detection.

Conclusions:

• The proposed Healing algorithm effectively corrects misclassifications in machine learning-based transport mode detection.
• The developed mobile application provides a highly accurate solution for identifying various transportation methods.
• This research offers a significant advancement for urban mobility analysis and transportation planning.