基于混沌时间序列的CPI短期预测分析

Analysis of CPI Short-term Prediction Based on Chaotic Time Series

有效控制CPI关乎国计民生,但其变动的随机性增加了预测难度。在对混沌时间序列预测流程进行了梳理的基础上,首先采用最大Lyapunov指数法辨别CPI时序的混沌特性,运用混沌理论重构相空间,利用逐步回归分析和BP神经网络进行混沌预测,并将ARIMA模型作为比较预测模型,最后从预测和拟合两个方面对模型进行效果评价。综合分析结果显示:2014年CPI增长范围为[2.8%,5.3%],变动幅度较大;2015年将高于4%;而2016年有望突破5%。该研究为CPI短期预测提供了较为可靠的方法,且预测结果可成为政府宏观调控政策的科学依据。

Effective control of CPI is vital to national economy and people＇s livelihood,yet its random movements increase the difficulty of predicting.Based on the summary the process of chaotic time series prediction,this paper first distinguishes the chaotic characteris-tics of CPI time sequence by using the method of the maximum Lyapunov index.Chaos theory is also utilized to reconstruct the phase space.Next,Chaos prediction is conducted under stepwise regression analysis and BP neural network,while ARIMA forecasting model is used as a comparison prediction model.Finally,the effect of the model is evaluated from both aspects of prediction and matching. The comprehensive analysis shows：CPI growth ranges from 2.8%to 5.3%in 2014,which is a considerable range.CPI will be higher than 4% in 2015 and hopefully exceeds 5%in 2016.The study provides a more reliable method for the short-term forecasts of CPI. The prediction results can be a scientific basis for government＇s implementation of various macro-control policies.