As traffic accident remains the leading causation of death in the world, this preventable tragedy could be better lessened by understanding the driver’s behavior further. In specific, which Driving Performance Measure (DPM) shall the research employ to evaluate the driving behavior and how the result should be interpreted requires more investigation. This is a twofold analysis to understand the DPM better: (i) 17 DPMs are calculated to explore its strength of association with the safety measures by comparing the odds ratio. (ii) the logistic regression model is employed to construct a multivariate linear prediction model. According to the result, in the one-to-one relationship, all the DPMs didn’t show a statistically significant relationship with the safety measure, either protective or risky. In terms of the logistic model, five DPMs (Mean Lateral Speed, Standard Deviation of Lateral Speed, Standard Deviation of Only Deceleration, Mean Acceleration Change Rate, Standard Deviation of Acceleration Change Rate) are selected as the predictor. Except for Standard Deviation of Only Deceleration, at the significance level of 0.05, the slope of the other four measures are significant in the regression formula. The model performs well in fitting the training data and reaches a precision of 60% in the test set. This study extends our recognition about DPM and can be considered as the reference for DPMs. Key words: driving performance measure, driving safety, naturalistic driving study, logistic regression model.

CICTP