Research On Bus Waitting Time Prediction Method Via Deep Learning

With the rapid growth of economy and the intensified development of urbanization,the living standard of residents has been constantly improved,and various kinds of transportation are becoming more and more convenient.The resulting traffic congestion,energy waste and environmental pollution have seriously affected the living quality of residents.Therefore,in order to reduce passengers' blind waiting time,increase passengers' travel efficiency,improve traffic conditions and transportation efficiency rate and utilization rate,this paper conducts an in-depth study and analysis on the waiting time required by passengers to take buses.The main research work is as follows:First of all,this paper puts forward the relevant research problems of bus waiting time prediction,discusses the research background and practical significance of this problem in detail,studies the development status of intelligent transportation system,summarizes the traditional traffic prediction model and method,and introduces the data set and data preprocessing work used in this paper in detail,including the processing of missing values,outliers and text data,the search of POI and the construction of waiting time data set.Finally,five traditional traffic prediction models are introduced.Secondly,with the analysis of the spatio-temporal characteristics that affect bus waiting time,it is found that the bus waiting time is a kind of time series data,which is related to both historical data of the same period and environmental impact of the outside world.Therefore,based on the advantages of LSTM nerual network in processing time series data,we establishe a bus waiting time prediction model based on the characteristics of spatio-temporal(LSTM-ST)by introducing the POI feature to depict spatial influencing factors and combining with temporal influencing factors such as weather.The experimental results show that compared with other traditional traffic prediction method,the LSTM-ST model proposed in this paper has lower prediction error and better model fitting effect.Furthermore,the study found that the bus waiting time data has obvious periodic characteristics in daily and weekly,so based on the above model,the attention mechanism is added and the periodic representation is introduced into the spatio-temporal prediction model.A bus waiting time prediction model based on periodic LSTM(LSTM-PE)is proposed,which effectively captures the day and week periodic characteristics of the bus waiting time data.Compared with other linear and nonlinear models,the LSTM-PE model further improves the prediction accuracy,and the prediction stability of the data sets sampled at different time intervals is better.