Outlier detection in data mining: Exclusion of errors or loss of information?

Purpose: Our research emphasizes the importance of considering outliers in production logistics tasks. With a growing amount of data, we require data mining to cope with these tasks. We underline that the widespread exclusion of outliers in data pre-processing for data mining leads to a loss of information and that using outlier interpretation can be used to address the issue. Methodology: The paper discusses the data pre-processing of data mining in production logistics problems. Methods of outlier interpretation are collected based on a literature review. In addition to the literature-based investigation, the work relies on a case study that illustrates the individual evaluation of outliers. Findings: This work shows that outliers take a special focus on the information generation. Within data pre-processing, a distinction must be made between an outlier as a defect and an outlier as a special datum. This can be conducted by methods presented in the literature. Originality: This paper adds to existing literature in the research field of insufficiently analyzed outlier interpretation and shows a need for research in data pre-processing of data mining.