Bus Arrival Time Prediction Based On GPS Data

Since the 21st century,the world is faced with the fundamental issue of energy depletion.As an important measure of national sustainable development,public transport priority strategy has been developed vigorously in more and more cities.At present,low service satisfaction of public transportation system is the main factor limiting its development.Bus arrival time information is the most concern for travelers,which directly affects the attraction of public transportation.Similarly,for bus managers,real-time and accurate bus arrival time information can provide a more scientific basis for scheduling.The advent of the era of big data and the development of data mining technology provides new research directions for bus arrival time prediction.Based on the actual GPS data of the bus in Hangzhou,this thesis studies the forecast of bus arrival time.First of all,the research significance of bus arrival time forecasting is elaborated and on the basis of summarizing the research results of domestic and foreign scholars at the present stage,the advantages and deficiencies of the existing research methods are pointd out in the thesis.Secondly,the bus GPS data mining technology is introduced and the data preprocessing is completed in this thesis.Considering the characteristics of bus operation,the idea of dividing bus arrival time into station stop time and station travel time is put forward,and the input data of relevant variables for model forecast are extracted accordingly.Based on the above analysis,four kinds of models are proposed,which includes historical data mean,time series,KNN algorithm and extreme learning machine(ELM),to predict bus arrival time.The result show that the four methods have their own characteristics in terms of prediction accuracy.Finally,considering the robustness of the forecasting process and results,interval analysis and traditional combined forecasting method are combined together.Meanwhile,based on the interval coefficient instead of the traditional combination forecast the weight of fixed point value,the interval combination forecast optimization model is established.The bus arrival time obtained is a number of intervals including the lower limit and the upper limit.The results show that the model is better than the single forecasting method in terms of the fitting degree of the forecasting results at various bus stations.