The data used in this study is the 5-min observations of 14 traffic weather monitoring stations located at the beginning section of G22 Qinglan Expressway and Jiaozhou Bay Bridge. The frequency and occurrence time of subfreezing temperatures (t＜0 ℃) in winter on the expressways around Jiaozhou Bay and their correlations with meteorological variables are statistically analyzed. The results are as follows. The risk of icy roads in winter is generally high in the study area, where the frequency of subfreezing road temperature is higher in the southwest than in the northeast. The diurnal variation of the hourly minimum road temperature follows the same pattern with the hourly minimum air temperature, and throughout the day, the minimum road temperature each hour is higher than the minimum air temperature. The air temperature and road temperature at 14:00, relative humidity at 18:00 are positively correlated with the next-day minimum road temperature, while the wind speed at 18:00 is negatively correlated with it. The prediction accuracy of t≤0 ℃ by the support vector machine (SVM) model with optimized parameters is 87.78%, higher than that by the multiple linear regression model. The improved performance is verified in the test of independence, which offers significant guidance for the application in icy road forecast service.