基于时间序列数据特性的缺失值估计算法

时间序列是在如运动捕捉、传感器网络、气候预报和财经市场预测等应用中的重要分析手段之一，然而在许多现实应用中经常发生观察数据缺失现象，如何应用相应的方法和模型来预测和填补含缺失数据的时间序列是目前研究的热点。以运动捕捉中遮挡问题为例提出了改进方法，利用平滑性和相互关联等时间序列数据特性，发现时间序列数据中的隐藏变量并挖掘它们的动态特性，在此基础上预测和填补时间序列的缺失值。实验结果证明了方法具有较小的数据重构误差，方法的计算时间应随着输入和运动捕捉持续时间增大而缓慢增长。

Missing value estimating algorithm based on time series data properties

Time series data are abundant in many application areas such as motion capture,sensor networks,weather forecasting,and financial market modeling.However,missing observations are hardly rare in these real applications,thus it remains a big challenge to model time series in the presence of missing data.With occlusion in motion capture as an example,a method is proposed to handle the challenge,which makes full use of temporal continuity and spatial correlation of time series data to identify hidden variables,to mine their dynamics,and to predict and recover missing values of time series.The experimental result shows that the approach can yield the best reconstruction error and the computation time grows slowly with the input and the time duration of the motion-capture.