Research on Vehicle Driving Conditions Based on Machine Learning

  • Chen Hua, Zheng Yuanyuan,Cai Guangxing


This paper takes the road driving data of light vehicles in Fujian Province as the research object, and collects 496464 sample data. First, we need to divide the kinematics segment and define 22 feature parameters, and use principal component analysis to reduce the dimension. Then, we analyze the clustering effect of the machine learning fuzzy C-means (FCM) algorithm and the K-Medoids algorithm. Secondly, we select the off-cluster Use the more effective FCM algorithm to construct the working conditions for the fragments near the center. Finally, we use the error analysis method to verify the model. The results show that comparing the experimental data with the characteristic parameters of the working condition model, the overall error is 9.18%, indicating that the working condition model constru5cted in this paper can represent the road conditions of Fujian Province well and has a good accuracy rate. On this basis, it is compared with the typical working conditions of foreign light vehicles. The results show that the foreign standard operating conditions are not suitable for the actual operating conditions of Fujian Province. It is recommended that the relevant departments of Fujian Province use their own actual road driving data to construct a more standardized and appropriate automobile working conditions and develop better automobile pollution control strategies.