Deep learning-based driver's driving propensity classification and identification method
In order to further investigate the dynamic change characteristics of drivers'driving propensity in the networked traffic environment,a deep learning-based driving propensity classification and identification method is proposed.Firstly,vehicle motion state parameters such as vehicle speed,acceleration standard deviation and headway time distance are extracted and a survey questionnaire is designed.A k-means clustering algorithm is used to initially classify driving propensity and correct vehicle state-related feature parameters,then a two-way long short-term memory neural network driving propensity recognition model considering attention mechanism is established to identify the driving propensity of surrounding vehicles.Finally,a control model is built to compare the recognition accuracy of the models.The results show that the recognition accuracy of the bi-directional long short-term memory network(BiLSTM)model considering the attention mechanism is 89.74％.Compared with the support vector machine(SVM)model and artificial neural network(ANN)model,the accuracy is further improved,which can realize the accurate and efficient recognition of dynamic driving tendency.
万方数据
维普
