Research On Prediction Of Conventio-Nal Public Transit Passengers’ Volu-Me In Small And Medium-Sized Cities Based On Neural Networks

With the accelerated process of urbanization in China, the number of small and medium-sized cities continues to increase,the total economy and the size of the population continues to expand in recent years. Development of public transit in small and medium-sized cities is closely related to the economy and convenience of residents’trips. Scientific and accurate public transit passengers’volume prediction is an important reference to planning and management of urban public transport.The concept of small and medium-sized cities and conventional public transit were defined firstly in this paper. Influencing factors of public transit passengers’volume in small and medium-sized cities were divided into external and internal factors to consider the effect of various factors respectively, which were found by studying development of traffic and analyzing the characteristics of public transit.The common methods that predicted public transit passengers’volume were summarized followed by this paper, such as time series, regression analysis, the gray model and neural networks. The theoretical basis and principle of these methods were elaborated and the advantages, disadvantages and applicability of each method were analyzed in predicting public transit passengers’volume. The conclusion that neural networks was suitable for public transit passengers’volume prediction was obtained in small and medium-sized cities.Then the limitation of BP neural network which was the most widely used in neural networks in forecasting and the improved ways were pointed out in the paper. The selection principle and method of predictor were put forward. The GA-BP network prediction model was built by combining neural networks and Genetic Algorithms. Genetic Algorithms and BP neural network were combined to define initial input weights and biases in BP neural network.The indicators concerning the development of public transit in a small and medium-sized city was cited finally. Gray correlation analysis method was used to determine the predictors. Effectiveness of GA-BP network model in prediction of public transit passengers’volume in small and medium-sized cities was tested through MATLAB programming examples. The applicability and superiority of model were proved by contrasted prediction results with single BP network model, exponential smoothing models, linear regression model and gray model in small and medium-sized cities.