Driver behavior profiling: An investigation with different smartphone sensors and machine learning.

Area of Science:

• Traffic Safety and Human Factors
• Machine Learning and Data Science
• Mobile Sensing Technologies

Background:

• Driver behavior significantly influences traffic safety, fuel consumption, and emissions.
• Driver behavior profiling aims to understand and improve driving habits through data analysis.
• Current methods often require specialized hardware, with a need for low-cost, high-performance solutions.

Purpose of the Study:

• To investigate the effectiveness of various Android smartphone sensors for driver behavior profiling.
• To evaluate different classification algorithms for driver aggressiveness classification.
• To identify optimal sensor-algorithm combinations for enhanced performance in driver profiling.

Main Methods:

• Utilized diverse sensors integrated into Android smartphones (e.g., accelerometer, gyroscope).
• Applied various machine learning classification algorithms to analyze driving data.
• Compared the performance of different sensor and algorithm assemblies for driver profiling.

Main Results:

• Certain combinations of smartphone sensors and intelligent classification methods demonstrated improved performance.
• Specific sensor data, when processed by advanced algorithms, effectively characterized driver aggressiveness.
• The study identified high-performing sensor-algorithm pairings for driver behavior analysis.

Conclusions:

• Android smartphone sensors, coupled with appropriate algorithms, offer a viable low-cost solution for driver behavior profiling.
• Optimized sensor-algorithm assemblies can achieve high performance in classifying driver aggressiveness.
• This research paves the way for more accessible and effective driver behavior monitoring systems.