The invention belongs to the road safety field and relates to a Bayesian network-based multi-data fusion algorithm. The algorithm involves linearity, traffic volume, climate environment, historical accidents, roadside features, roadside traffic accident occurrence probability, and monitoring opinions, wherein the linearity includes an average curve and a longitudinal gradient, the traffic volume includes daily average traffic volume and the proportion of trucks, the climatic environment includes the frequencies of rain, snow and fog climate and the frequencies of other natural disasters, the historical accidents include wounding accidents, fatal accidents and property loss accidents, the roadside features include roadside depth, discrete hazardous materials, continuous hazardous materialsand clear zone conditions, and the monitoring opinions are variables. According to the algorithm of the present invention, the Bayesian network, serving as a reliability analysis means, processes complex logic relationships, and therefore, multi-state variables and related lines between the variables can be processed conveniently, and uncertainty relationships between the variables can be well expressed. The algorithm can have high real-time performance and is scientific and reasonable.