Performance Comparison of Two Statistical Parametric Methods for Outlier Detection and Correction

Abstract

Abstract Outlier detection and correction referred to as data preprocessing, is crucial in time series analysis and modeling. It has been a challenge to preprocess a volatile time series data possessing intricate trend characteristics. Two well-established statistical parametric methods, such as improved sliding window prediction and portrait dataset-based, perform adequate data preprocessing. While the former is equipped with an optimal window width selection approach, the latter, on the other hand, is based on a data visualization approach named portrait. This paper compares both methods’ preprocessing performance when applied to seasonal time series data with varying time resolutions and complex trend patterns for different content of outliers through detailed result analyses. Further, a new metric to measure outlier correction capability is suggested.