| The Consumer Price Index,or CPI,is an important indicator used to reflect the price index of consumer goods and services.This index directly affects the purchasing power of the people and the stable development of the macro economy.Therefore,the growth and fluctuations in the CPI index are particularly important for research and forecasting.This paper is based on the CPI index data from January 2008 to December 2017.Based on the traditional BP neural network,this paper compares different learning functions to get the most suitable learning function for this article.At the same time,it improves the network learning,including increasing momentum terms and adjust learning rate,improve the stability of the model,and finally construct an improved BP neural network model.Using the trained neural network to predict the CPI index for the next 12 months,the average error rate is only 0.4%.At the same time,using the method of this paper,substitute the CPI index data in the same type of literature,and through comparative analysis,the improved BP neural network is better than BP.The neural network results are much more accurate,which fully shows that the neural network has great application space in predicting the CPI index,and with the improvement of the learning algorithm,the accuracy of the CPI index will be higher and higher. |
