Driver behavior profiling: An investigation with different smartphone sensors and machine learning
DOI: 10.1371/journal.pone.0174959
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of driver behavior profiling, a process aimed at improving traffic safety, reducing fuel consumption, and lowering emissions by characterizing driver aggressiveness. While previous studies have utilized smartphone sensors or dedicated telematics devices to monitor driving, there is a lack of quantitative assessment comparing different combinations of smartphone sensors and machine learning algorithms (MLAs) in real-world scenarios. The authors aim to identify the optimal assembly of sensors, sensor axes, MLAs, and sliding window sizes to detect aggressive driving events with high performance using low-cost Android smartphones. The study employs a multi-label supervised learning classification framework. The researchers collected data from four Android smartphone motion sensors: accelerometer, linear acceleration, magnetometer, and gyroscope. Raw sensor data were sampled and translated from the device coordinate system to Earth’s coordinate system to ensure independence from the phone’s position within the vehicle. This translated data was used to generate attribute vector datasets. The authors evaluated four distinct MLAs—Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forest (RF), and Bayesian Networks (BN)—chosen to represent diverse algorithmic approaches. Performance was measured using the Area Under the Receiver Operating Characteristic curve (AUC), where values closer to 1.0 indicate superior detection capability. The experimental design involved systematically varying elements of the evaluation assembly, including sensor type, axis selection, MLA configuration, and the number of frames in the sliding window, to determine the best match for each specific driving event class. The investigation revealed that specific combinations of sensors and intelligent methods significantly improve classification performance compared to isolated approaches. The results demonstrate that no single sensor or algorithm is universally optimal; rather, the effectiveness of detection depends on the specific type of aggressive driving event being monitored. By evaluating various assemblies, the study identified configurations that maximize AUC scores for different maneuvers, confirming that sensor fusion and tailored machine learning models are critical for accurate profiling. The findings validate that smartphone-based sensing, when paired with appropriate MLAs, can effectively detect aggressive driving behaviors such as harsh acceleration, braking, and turning. The significance of this work lies in its contribution to the development of low-cost, high-performance driver monitoring systems. By establishing which sensor and algorithm combinations yield the best results, the study provides a foundation for more effective Usage-Based Insurance (UBI) and freight management applications. These systems can reward safe driving, reduce accidents, and improve resource economy by providing real-time feedback to drivers. The research underscores the potential of leveraging ubiquitous smartphone technology and diverse machine learning techniques to create scalable solutions for traffic safety and energy efficiency, moving beyond the threshold-based or limited algorithmic approaches prevalent in prior literature.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | OpenAlex-citations | — | — | 1 | 2026-06-25 |
| archive | success | unpaywall | — | — | 2 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Empirical Findings: observational prevalence
- Methodological Resource: tool software
- Theoretical Contribution: computational model