Identification of driving simulator sessions of depressed drivers: A comparison between aggregated and time-series classification
DOI: 10.1016/j.trf.2020.09.015
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This study addresses the gap in transportation research regarding the proactive identification of depression in drivers using microscopic driving data. While previous studies have established that depression increases collision risk and deteriorates driving performance, they typically rely on aggregated data and simple statistical methods to compare pre-diagnosed groups. This paper aims to reverse this approach by utilizing machine learning to detect whether driving simulator sessions belong to depressed drivers based solely on kinematic data, comparing the efficacy of aggregated data against time-series classification. The researchers utilized data from a driving simulator experiment involving 11 older drivers with depressive symptoms and 65 healthy controls. The dataset comprised highly disaggregated measurements of speed, lateral acceleration, and longitudinal acceleration, recorded every 17–33 milliseconds. To address the significant class imbalance (ratios ranging from 1:6 to 1:7), the study employed Random Forest classifiers combined with the SMOTE-ENN imbalanced learning technique, which synthesizes minority class samples and cleans the dataset. The analysis compared two methodological frameworks: one using data aggregated into 30-second, 1-minute, and 5-minute intervals, and another using corresponding time-series of these kinematic variables. Correlated safety indicators like Time-to-Collision were excluded to prevent multicollarity, focusing the model on mean and standard deviation metrics of speed and acceleration. The results demonstrated that machine learning classifiers could effectively distinguish depressed driving sessions from healthy ones with high accuracy and low false alarm rates. Crucially, classifiers trained on time-series data outperformed those using aggregated statistics. Specifically, time-series of mean speed and the standard deviation of longitudinal acceleration over short durations (30 seconds to 1 minute) proved to be the strongest predictors. For instance, the classifier using 30-second time-series of mean speed achieved a recall of 99.81% and a false alarm rate of 14.96%. Models without imbalanced learning techniques failed to identify depressed drivers, highlighting the necessity of handling data scarcity. Furthermore, shorter time-series durations yielded better predictive power than longer intervals, suggesting that rapid fluctuations in speed and acceleration are key indicators of depressive symptoms. The significance of this work lies in its validation of microscopic, real-time data for detecting cognitive impairments, moving beyond post-hoc aggregated analysis. The findings suggest that Intelligent Transportation Systems could potentially use such classifiers to proactively identify at-risk drivers, enhancing traffic safety. The study concludes that naturalistic driving data and advanced techniques like deep learning may further improve detection efficiency, offering a pathway for integrating mental health monitoring into autonomous driving and driver assistance systems.
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 | Crossref | — | — | 1 | 2026-06-07 |
| archive | success | canonical_url | — | — | 7 | 2026-06-09 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-09 |
| chunk | success | chunk | — | — | 1 | 2026-06-09 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-09 |
| enrich | success | semantic_scholar | — | — | 1 | 2026-06-10 |
| promote | success | — | — | — | 1 | 2026-06-07 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 8 | 2026-06-11 |
| verify | success | — | — | — | 1 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; 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).
- Methodological Resource: validation psychometrics, tool software
- Theoretical Contribution: computational model