Formation Mechanisms and Clustering Differences in Risky Riding Behaviors of Electric Bike Riders

Tao Wang, Yuzhi Chen, Jin Yu, Sihong XieIEEE Access 2021 · OpenReview

Abstract

Individual differences among electric bike (e-bike) riders significantly contribute to risky behaviors such as traffic violations, negligence, and pushing limits, leading to high accident rates. This study analyzes 573 riders to identify subcluster characteristics based on riding confidence, risk perception, and safety attitudes. Using the E-bike Risky Riding Behavior Questionnaire (E-RBQ), factor analysis, and K-means clustering, riders were categorized into four types: Action, Anxiety, Introversion, and Negative. A structural equation model (E-SEM) was established to explore the formation mechanisms of risky behaviors for each type. The findings suggest that overconfidence, poor risk awareness, and herd mentality are key drivers, necessitating tailored safety interventions for different rider profiles.

Introduction

Electric bikes have become a global success for sustainable urban mobility, yet they are frequently involved in road accidents. While external factors like road conditions are often studied, rider heterogeneity—the psychological differences between individuals—is frequently ignored. Treating all e-bike riders as a monolithic group fails to capture why one rider might speed through a red light while another makes a dangerous turn out of hesitation. This research bridges this gap by modeling how individual psychological traits manifest as specific risky behaviors.

Methodology

The study utilized a multi-stage analytical approach:

  1. Survey Design: The E-bike Risky Riding Behavior Questionnaire (E-RBQ) was developed, featuring 81 items to measure confidence, risk perception, and safety attitudes.
  2. Data Collection: 573 valid questionnaires were collected from riders in Guilin and Nanning, China.
  3. Factor Analysis: Data was distilled into 12 factors, including technical ability, judgment ability, and crowd psychology.
  4. Clustering: K-means clustering identified four distinct rider types based on the Calinski-Harabaz index.
  5. Structural Equation Modeling (SEM): An "E-SEM" model mapped causal pathways between psychological traits and four types of risky behaviors: negligence/error, violation, pushing limits, and "taking the lead."

Results and Discussion

The analysis revealed a clear taxonomy of e-bike riders, each driven by different psychological mechanisms as shown in the structural relationship model below:

Figure 6: Structural relationships between rider clusters and psychological factors (A: Riding Confidence, B: Risk Perception, C: Safe Attitude).
  • Action Type (n=152): Characterized by high confidence but low risk perception. Their negligence and errors (0.48) and "take the lead" behaviors (0.44) are significantly and positively influenced by their judgment ability.
  • Anxiety Type (n=139): Possess medium confidence with very low risk awareness. Their violations (-0.52) and "take the lead" behaviors (-0.41) are significantly and negatively influenced by their knowledge of traffic rules.
  • Introversion Type (n=160): High risk perception but extremely low confidence. Their pushing of limits (-0.29) and "take the lead" behaviors (-0.31) are significantly and negatively influenced by their probability evaluation of danger.
  • Negative Type (n=122): Low safety attitude and high susceptibility to "herd mentality." This crowd psychology significantly influences their negligence (-0.43) and violations (0.37).

The SEM results highlight that confidence is a double-edged sword. While high confidence in Action types leads to miscalculating traffic gaps, the lack of confidence in Introversion types leads to equally dangerous errors through hesitation and over-correction.

Conclusion

Tailored traffic safety interventions are more effective than "one-size-fits-all" regulations. Action and Anxiety types require education to temper overconfidence in their technical skills and judgment, while Negative types benefit from measures that discourage following the "herd" in violating traffic rules. Future work should aim to integrate real-world trajectory data with these psychological profiles to develop real-time safety warning systems.

Summary updated

Abstract

Individual differences between various riders cause risky riding behaviors such as violations, taking the lead, negligence and error, and pushing the limits, resulting in a high incidence and high number of road accidents for the vulnerable road use group of electric bike riders. Therefore, the subcluster differential characteristics among riders were analyzed in terms of their riding confidence, risk perception, safety attitude, and basic attributes. The influences and formation mechanisms of risky riding behaviors among the subclusters of riders were also explored. First, the 573 riders were clustered into 4 types, action type, anxiety type, introversion type, and negative type, based on the E-bike Risky Riding Behavior Questionnaire (E-RBQ), factor analysis method, and K-means clustering. Second, a structural equation model of e-bike risky riding behavior (E-SEM) was established to explore the main influencing factors for the risky riding behavior of the 4 types of riders and the differences among them. Finally, risky riding behavior avoidance strategies for various types of riders were proposed. The findings showed that negligence and error (0.48) and take the lead behavior (0.44) of action types were significantly and positively influenced by judgment ability; violation behavior (-0.52) and take the lead behavior (-0.41) of anxiety types were significantly and negatively influenced by traffic rules; pushing the limits (-0.29) and take the lead behaviors (-0.31) of introversion types were significantly and negatively influenced by probability evaluation; and negligence and error (-0.43) and violation (0.37) of negative types were negatively and positively influenced by herd mentality. In particular, the overconfidence of the action and anxiety types in their own techniques and judgment ability may cause misjudgment of the surrounding area; the worry degree of the introversion type must be balanced effectively; and the negative type must control the degree of confidence in their judgment ability.
BibTeX
@inproceedings{wang2021formation,
  author = {Tao Wang and Yuzhi Chen and Jin Yu and Sihong Xie},
  title = {{Formation Mechanisms and Clustering Differences in Risky Riding Behaviors of Electric Bike Riders}},
  booktitle = {IEEE Access},
  year = {2021},
  url = {https://openreview.net/forum?id=HHRP9IURgY},
}
Figure