为能同时检测时间序列中的附加异常和革新异常,改进自回归模型,提出距离因子递推最小二乘(DF-RLS)线性预测算法。在此基础上,给出一种基于距离和DF-RLS的联合异常检测方法——DDR-OD。实验结果表明,与当前其他时间序列异常检测方法相比,DDR-OD的检测效果较优。
In order to detect both Additive Outliers(AO) and Innovation Outliers(IO) in time series,this paper improves the linear prediction of time series,proposes a Distance Factor Recursive Least Square(DF-RLS) algorithm.It combines DF-RLS with distance-based outlier detection method,proposes a time series outlier detection method based on distance and DF-RLS,named DDR-OD.Experimental results show that the DDR-OD is an effective method for time series outlier detection.