Research on Classification Model of Natural Driving Scenario Based on LightGBM

With the rapid development of Artificial Intelligence and big data analysis, the application of driving scenarios has become a new hot spot in the research of smart connected vehicles. This study uses the collected Controller Area Network (CAN) bus and CARTAC driving scenario standard marker data implemented Discrete Wavelet Transform (DWT) and LightGBM algorithm, builded a high-speed driving scenario automation classification model to achieve follow-up driving, patrol Line driving, lane changing, adjacent car cutting in and front car cutting out five typical scenarios automatic classification. The accuracy of classification of all samples is 83%, and the accuracy rate of sub-scenario classification is over 78%. Compared with other supervised learning models, it has better automatic classification effect, effectively solving the problem of automatic processing of scenario data facing the industry. Industrial application value.