Urban Mobility Pattern Detection: Development of a Classification Algorithm Based on Machine Learning and GPS.

Area of Science:

• Computer Science
• Urban Planning
• Transportation Engineering

Background:

• Accurate transportation mode classification is crucial for effective urban planning and traffic management.
• Existing methods may lack the granularity or accessibility needed for real-time urban mobility analysis.
• Smartphone sensors offer a ubiquitous platform for collecting detailed mobility data.

Purpose of the Study:

• To develop and validate an innovative algorithm for classifying diverse transportation modes using smartphone sensor data.
• To extract significant mobility pattern features for enhanced transport classification accuracy.
• To lay the groundwork for generating origin-destination matrices for improved urban mobility understanding.

Main Methods:

• Collected transportation data (date, time, GPS, altitude, speed) via a dedicated smartphone application.
• Analyzed stopping times, distance, and average speed to identify distinctive transport mode patterns.
• Employed a decision tree model trained on extracted features like speed, acceleration, and longitudinal dynamics.

Main Results:

• Achieved high classification accuracy: 94.6% in validation and 94.9% in testing.
• Demonstrated strong model performance with precision (0.8938), recall (0.83084), and F1-score (0.86117).
• Successfully classified modes including walking, biking, tram, bus, taxi, and private vehicles.

Conclusions:

• The developed algorithm effectively classifies transportation modes using smartphone data, offering a reliable tool for urban mobility analysis.
• The method has significant potential applications in urban planning, transport management, and public transport optimization.
• This research provides a foundational step towards creating origin-destination matrices for a deeper understanding of urban movement.