Research On Traffic Mode Recognition Based On Convolutional Neural Network

The choice of transportation mode greatly affects transportation planning,which is the core of transportation planning and traffic management.In addition,user behavior can be discovered according to the user's transportation mode,which is used to provide personalized recommendation and intelligent service.Using the user's GPS trajectory to mine its mobile information has become a hot research topic.The key to understanding the user's mobility is to identify transportation mode from the trajectory.At present,the machine learning method based on artificial feature extraction for transportation mode recognition has the problem that feature extraction is vulnerable to environmental conditions,human bias.And excessive extraction of features will face the challenge of a large number of dimensional descent.Convolutional neural networks,which have been widely used in image processing and speech recognition,have the advantages of automatic feature extraction and network weight sharing.Given the above considerations,this paper uses the convolutional neural network method to identify the transportation mode of the moving trajectory.First,the sample size is increased by the segmentation process in the preprocessing.Then the seven different setting conditions of the model are trained,using the trajectory motion properties such as velocity,acceleration,jerk and bearing rate as the input.Finally,the optimal segmentation value and the number of iterations are evaluated through multiple experiments to achieve the best architecture for identifying five modes of transportation,including walking,cycling,bus,driving and train.A simple trajectory identification system is constructed.In order to prove the superiority of the model,it is compared with the classical RF,DT and other machine learning methods.The experimental results certify the advantages of the CNN model.Moreover,it is also found that the number of iterations and the value of the segmentation value have a great influence on the experimental accuracy and running time.The optimized model improved the accuracy of the convolutional neural network model by4.2%,eventually reaching 83.7%,and the running time of which is reduced.