PyODDS: An End-to-End Outlier Detection System

Department of Computer Science and Engineering Texas A&M University College Station, TX 77840, USA Abstract PyODDS is an end-to-end Py thon system for O utlier D etection with Database Support. It provides various outlier detection algorithms which meet the demands for users in different fields, with or without data science or machine learning background. PyODDS gives the ability to execute machine learning algorithms in-database without moving data out of the database server or over the network. It also provides access to a wide range of outlier detection algorithms, including statistical analysis and more recent deep learning based approaches. Keywords: anomaly detection, end-to-end system, outlier detection, deep learning, machine learning, data mining, full stack system, data visualization 1. Introduction Outliers refer to the objects with patterns or behaviors that are significantly rare and different with the rest of majorities.