时间序列分析在粮食产量预测中的应用

摘要

Considering the practical application,we propose a new Bayesian estimation procedure based on prior information with the analysis of time series by Box-Jenkins method. Since the complicated computation of posterior probabilities is involved in the estimation method proposed above,a new algorithm concerning how to compute the posterior probabilities is designed based on Gibbs and SIR algorithms. Moreover, a real time series of grain yield production from 1949 to 2009 were analyzed, and make a prediction of 2010'grain yield production.%出于对实际应用的考虑,在利用Box-Jenkins建模方法处理时间序列数据时,给出了一种新的基于先验信息的Bayes估值方法.由于此估值方法涉及到复杂的后验概率计算问题,所以本文给出了运用Gibbs抽样及SIR抽样算法来对后验概率进行计算的具体算法.其次,应用以上理论,对1949年至2009年的粮食产量数据进行分析,并对2010年的粮食产量数据进行了预测.