Design And Implementation Of A Traffic Accident Prediction System Based On Machine Learning

With the development of society,the transportation facilities are more developed,and people's travel needs are also increasing,but the traffic safety problems that follow are also increasing.Traditional vehicle accident prevention measures include passive safety systems and active safety systems.In recent years,with the advancement of technology,a new generation of safety systems such as assisted driving has slowly entered the stage,but according to research,a more reasonable safety system is bound to be Preventive safety system based on vehicle networking technology and accident prediction technology,because the vehicle networking technology enables vehicles and vehicles to communicate under non-line-of-sight conditions,solving the problem of limited driver's vision,and accident prediction can determine whether it is dangerous in advance.To better protect passenger safety and reduce losses.Traditional accident prediction often applies the classical prediction model to a small set of data,so that the obtained model can only predict a small range of traffic accidents or traffic accident probabilities,and most accident prediction models are limited by fewer data features.Or the actual application scenario is not considered,and the characteristics of the accident impact are not classified,which does not reflect the practical significance of the accident prediction model.Based on the consideration of the above issues,this paper has done the following work:1.The paper carried out preliminary data processing on the vehicle accident big data set,and analyzed the relationship between traffic accident factor variables and accidents from five aspects: human,vehicle,road,environment and other aspects,and carried out the traffic accident factor variables.Classification and re-encoding,paving the way for the feature processing of traffic accident prediction.2.The original data and the heterogeneous spatial data are processed separately,and the logistic regression and random forest algorithm are applied to the two sets of data and their eigenvalues.After the analysis and training,the corresponding prediction model is obtained.After the experimental analysis,the data is different.The prediction model is applied after the quality space is processed.The prediction model obtained by this method has better accuracy than the prediction result obtained by using the prediction model for the original data.3.Using SUMO,VEINS and OMNET++ software to build the vehicle network simulation environment,and design a preventive safety system based on accident prediction in the simulation environment,and analyze and demonstrate some key parameters.The system can warn the driver in dangerous situations.To avoid potential traffic accidents,and finally summarize and propose corresponding recommendations for road safety.