Affiliation:
1. Bengal College of Engineering and Technology, India
2. Dept of CSD, Dr. B.C. Roy Engineering College, Durgapur, India
Abstract
Throughout many real-world investigations, outliers are prevalent. Even a few aberrant data points can cause modeling misspecification, biased parameter estimate, and poor forecasting. Outliers in a time series are typically created at unknown moments in time by dynamic intervention models. As a result, recognizing outliers is the starting point for every statistical investigation. Outlier detection has attracted significant attention in a variety of domains, most notably machine learning and artificial intelligence. Anomalies are classified as strong outliers into point, contextual, and collective outliers. The most significant difficulties in outlier detection include the narrow boundary between remote sites and natural areas, the propensity of fresh data and noise to resemble genuine data, unlabeled datasets, and varying interpretations of outliers in different applications.
Reference36 articles.
1. Enhancing density-based clustering: Parameter reduction and outlier detection
2. DBOD-DS: Distance-based outlier detection for data streams.;MShiblee Sadik;Database Expert Syst Appl.,2010
3. RS-Forest: A rapid density estimator for streaming anomaly detection.;KWu;IEEE Int Conf Data Min.,2014
4. Outlier detection in streaming data: a research perspective.;NChugh;2014 Int Conf Parall Distrib Grid Comput,2014
5. Fast and scalable outlier detection with approximate nearest neighbor ensembles.;ESchubert;Database Syst Adv Appl.,2015