Abstract:Based on the daily traffic accidents data and the observed meteorological data in Nanjing region in 2012, the automatic regression (AR) model of the daily traffic accidents amount was established by multi-factor time series analysis, taking autocorrelation into consideration. Analysis on the data of weekdays and weekends showed that the relationship between adverse weather condition and daily traffic accident amount in weekdays was more significant than it was in weakends. Precipitation, low temperature, low visibility were significantly correlated with the traffic accident. Daily precipitation below moderate intensity was positively related with daily traffic accident amount. The temperature between 2-12 ℃ and the minimum visibility between 200-500 m were negatively correlated with daily traffic accident amount. However, the larger precipitation intensity, the lower temperature and the minimum visibility were all slightly correlated with daily traffic accident amount. Furthermore, a stepwise regression (SR) model was established on the basis of different influences of meteorological factors on the daily traffic accident amount. Finally, the performance and results by the test of fitting goodness were compared between these two models, showing the AR model performs better than the SR model in weekdays.