Public Transport Passenger Flow Analysis And Short-term Prediction Research Based On Data Mining

To solve the prominent contradiction of urban traffic supply and demand, improve the level of urban life, promote the city social and economic sustainable development, we must increase the pace of the urban traffic construction, planning and management. At the same time, take various means to control urban traffic structure, guide the urban traffic to take public transport as the main body in the direction of development. City transportation development experience and lessons at home and abroad proves that the priority to the development of public transportation is the basic way to solve the urban traffic. Bus passenger flow data is the foundation of the bus operation scheduling optimization. Accurately to make bus routes prediction can effectively guide the bus operating decisions, formulate operation scheduling scheme, effectively improve the operating efficiency of bus system.Along with the development of the intelligent bus dispatch system, the bus company has accumulated a large number of bus IC card, GPS, car video and other data, the traditional statistical analysis method, already could not satisfy the present stage, bus passenger flow analysis and short-term prediction of real-time requirement, the paper’s innovative use of the technology of data mining, to extract the key features of the bus IC card and GPS data, correlation analysis, on the basis of traditional prediction algorithm in dealing with noise data such as low precision, slow convergence speed, based on the extreme learning machine bus passenger flow of short-term forecasting model, based on the model parameter of the weighted processing, reduce the sensitivity to noise data, through the experimental analysis of the three prediction algorithm and machine learning algorithm of the limit, in the aspect of process noise data fastest convergence speed and prediction accuracy of the highest.This article mainly work as follows:(1) Based on data mining technology, IC card, GPS, transit site information, such as data fusion, to create the data warehouse, extract the key attributes, fully association rules analysis, get the characteristics of passenger flow distribution (including time features, spatial characteristics, short).(2) Because the models such as bayes and support vector machine (SVM) in terms of short-term passenger flow prediction algorithm is iterative cumbersome, slow convergence speed, and sensitivity to noise data, bus passenger flow prediction model based on extreme learning machine is established, on the basis of the existing data mining, model for two kinds of weighted processing, reduce the influence of the noise data of the model. Experimental analysis show that the limit of machine learning algorithm than the naive baves and support vector machine forecasting precision is higher.