Multi-step histogram based outlier scores for unsupervised anomaly detection: ArcelorMittal engineering dataset case of study

Anomaly detection is the task of detecting samples that behave differently from rest data or include abnormal values. Unsupervised anomaly most common scenario, which implies algorithms cannot train with a labeled input and do not know behavior beforehand. Histogram-based methods are one approaches in unsupervised detection, remarking good performance low runtime. Despite performance, histogram-based detectors capable processing flows while updating their knowledge deal high amount samples. In this paper, we propose new approach for addressing aforementioned problems by introducing ability to update information inside histogram. We have applied these strategies design algorithm called Multi-step Histogram Based Outlier Scores (MHBOS), including five histogram mechanisms. The results shown validity MHBOS as well proposed terms computing times.