基于可逆跳MCMC的AR模型阶次判定及其在陀螺漂移预测中的应用

为消除传统方法在样本数较少时确定模型的不足,提出了一种在贝叶斯准则下将可逆跳MCMC法用于AR模型的阶次估计,以抽样的方法解决AR模型对时间序列拟合时的阶数不确定问题。给出了阶次估计算法的公式和步骤。仿真试验和某陀螺漂移模型估计的结果表明,该法预测结果与实际较为吻合。但为进一步提高预测精度,还需研究平稳性和初始状态的影响。

Model Order Selection of Autoregressive Process Based on the Reversible Jump MCMC and Its Application on Gyro''''s Drift Prediction

To overcome the disadvantages of classic method to predict AR model with small species,the reversible jump Markov chain Monte Carlo(MCMC) was put forward to estimate the order for AR model under Bayesian selection in this paper. The order uncertainness of AR model in fitting a time sequence was solved by sampling.The equation and steps of the order estimation algorithm were presented. The simulation result and estimation for some gyro drift showed that prediction of this method was agreed with the data set.But to improve the accuracy of the prediction,the effect of the smooth and initial state should be studied more.