Spatial analysis of telematics-based surrogate safety measures

Area of Science:

• Transportation Science
• Road Safety Engineering
• Spatial Data Analysis

Background:

• Surrogate Safety Measures (SSMs) like harsh braking are valuable for road safety analysis when crash data is scarce.
• Spatial analysis of harsh braking events is underexplored, particularly in regions with low crash frequencies.

Purpose of the Study:

• To conduct a spatial analysis of harsh braking events to assess their adaptability and informativeness for road safety.
• To explore the correlation between geometrical/behavioral parameters and harsh braking events using advanced statistical and machine learning models.

Main Methods:

• Utilized smartphone driving data and OpenStreetMap road network data for 6,103 road segments.
• Applied advanced statistical and machine learning models, including spatial autocorrelation analysis.
• Developed and compared spatial models (Spatial Zero-Inflated Negative Binomial, Spatial Random Forest) against non-spatial counterparts.

Main Results:

• Harsh braking events positively correlate with trip frequency, segment length, speeding, and mobile phone use.
• Motorways showed fewer harsh braking events; trip frequency was the most influential predictor of risk exposure.
• Spatial models (SZINB, SRF) demonstrated superior fit and reduced spatial autocorrelation compared to non-spatial models.

Conclusions:

• Combining geometrical and behavioral parameters with spatial analysis offers a proactive approach to road safety.
• The study validates the effectiveness of spatial modeling techniques (SZINB, SRF) for analyzing harsh braking events.
• Findings can inform policymakers in developing targeted countermeasures to improve road safety and reduce harsh braking incidents.