Abstract: This article aims to study the impact of precipitation on driving travel time and its forecasting model.The current road condition value,precipitation amount,historical road condition changing trend,and current driving navigation travel time are used as independent variables,and the increase rate of driving navigation travel time in the next hour is used as the dependent variable.A total of 13 142 pairs of meteorological and traffic data from January 2021 to December 2022 in the jurisdiction of Fuzhou City were collected through web crawlers to analyze the impact of precipitation on driving travel time.Multiple linear regression and random forest regression were used to develop two forecasting models,and the prediction effects of the two models were tested using the Fuzhou City jurisdiction from January to May 2023 as an example.The results indicate that precipitation has a non-linear positive impact on driving travel time.Within the rainfall intensity range of 0.0-0.7 mm·h^-1,for every 1mm increase in rainfall,the average travel time increases by 0.33 times.Beyond 0.7 mm,for every 1mm increase in rainfall,the average travel time increases by 0.06 times.The rate of increase in travel time varies slightly across different time periods,with a higher increase during commuting hours.Whether using forecast precipitation data or actual precipitation data,the random forests have lower losses or biases in sample training and testing compared with the multiple linear regression models; and the accuracy of precipitation forecasts has a significant impact on travel time forecasting.

Key words: Road conditions, Travel time, Impact forecasting, Big data mining