Forecast Of Railway Passenger Volume Based On Improved Neural Network And Its Time Series Combination Model

With the rapid development of Chinese economy and the continuous improvement of people's living standards,Chinese railway industry has also made an unprecedented development and prosperity,but it still faces challenges.How to adapt the railway transportation capacity to the needs of the national economy and social development is the main difficulty which Chinese railway industry is facing.Railway passenger volume has a direct impact on railway transportation dispatching,so how to accurately predict the railway passenger volume is particularly important.Due to the non-linear,time-varying and uncertain nature of railway passenger volume data,it is difficult to accurately describe its objective laws.Different prediction methods have different prediction accuracy,a single prediction method usually has a certain scope of application and restrictions,and it is prone to the problem of unstable prediction results.Bates and Granger proposed that the combined predictive performance of the combined prediction model is better than the single-item prediction model,because a single model can only reflect part of the information and the change law of the data,but the combined prediction model can make up for this shortcoming,it can improve the effect of overall prediction by comprehensively using multiple prediction methods to make combined predictions.On the basis of full considering the characteristics of railway passenger volume data,this paper introduces probability weights in the combined prediction model,determines the lower limit of weights based on Chebyshev's theorem,and constructs a combination prediction model of probability weights which is made up of BP neural network model,GM(1,1)model and ARIMA model;then improves the BP neural network model to improve the overall prediction accuracy of the combined model.The main work of this article is as follows:(1)In view of the uncertainty of railway passenger volume data,it is necessary to combine multiple different models into a new model to predict railway passenger volume.According to the different characteristics of the data,the following single models are selected: according to the data,it has a form similar to time series,select the ARIMA model;choose the GM(1,1)model based on the number of data samples;choose the BP neural network model based on the nonlinearity of the data.(2)Based on the homogeneity of probability and weight coefficient,determine the optimal weight coefficient of a single model,and establish a combination model to prove the effectiveness of the probability weight method.Because the prediction accuracy of a combination model mainly depends on the merits of each single model,the BP neural network model is improved by increasing the number of hidden layers,replacing Sigmoid with Softplus as the activation function to improve its prediction accuracy,and obtaining a new combined prediction model.Then the equal weighted average method and the inverse variance method are selected respectively.And entropy weight methods,establish a combination model for horizontal comparison.In this paper,the mean absolute percentage error(MAPE)and the mean square error(MSE)are used as evaluation indexes to compare the prediction results of each model.It is found that the combination model proposed in this paper has higher prediction accuracy,smaller fluctuations,and more reliable results.The effectiveness and practicability of the combined model can provide an effective reference for the relevant departments to predict railway passenger volume.