A hybrid feedforward neural network algorithm for detecting outliers in non-stationary multivariate time series

作者：

Highlights：

• The presence of outliers can lead to model misspecification and poor forecasts.

• A new outlier detection algorithm using feedforward neural network is given.

• The proposed algorithm does not require a priori knowledge of model parameters.

• The proposed algorithm performs well for detecting outliers in small & large clusters.

摘要

•The presence of outliers can lead to model misspecification and poor forecasts.•A new outlier detection algorithm using feedforward neural network is given.•The proposed algorithm does not require a priori knowledge of model parameters.•The proposed algorithm performs well for detecting outliers in small & large clusters.

论文关键词：Feedforward neural network,Mahalanobis distance,Non-stationary time series,Outlier detection,Robust estimate

论文评审过程：Received 10 March 2020, Revised 21 June 2021, Accepted 1 July 2021, Available online 7 July 2021, Version of Record 14 July 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115545