ARIMA时间序列模型和BP神经网络组合预测在铁路客座率中的应用

Application of ARIMA Time Series and BP Neural Network Combination Model in Railway Passenger Rate

首先利用python3.5对铁路客座率原始数据进行预处理,然后利用ARIMA时间序列和BP神经网络进行单一的模型预测,得出单一预测模型的均方误差.在组合预测求解时,先求出ARIMA时间序列模型的误差向量E1和BP神经网络的预测误差为E2,由于这两种预测方法是相互独立的,因此误差向量E1和E2线性无关且组合预测误差向量为E=（E1,E2）,得出组合预测平方和的形式为J-W^TEW,然后根据组合预测误差平方和最小的原则来确定权值w1,w2,最后求解凸二次规划问题得到权值并求出组合预测模型和均方误差.通过比较单一模型预测和组合预测的均方误差,得出结论：组合预测模型的精确度高于单一预测模型的精确度.

Firstly, using python3.5 to preprocess the original data of railway passenger rate,then using ARIMA time series model and BP neural network model to forecast, the mean square error of a single prediction model is obtained. In combination forecasting, the error vector of ARIMA time series model and BP neural network model are obtained. Because the two methods are independent of each other, and error vector are independence. It is concluded that the sum of squares of combination forecasting is, then, according to the principle of minimum square sum of combination forecasting error, the weights are determined.Finally, the convex quadratic programming problem is obtained and the combination forecasting model and mean square error are obtained… by comparing the mean square error of single model prediction and combination forecasting, come to conclusion： the accuracy of the combined forecasting model is higher than that of the single prediction model.