TODS: An Automated Time Series Outlier Detection System

TL;DR

This paper introduces TODS, an automated, modular system for time-series outlier detection that assembles end-to-end pipelines from data processing to detection using a library of 70 primitives. Pipelines are modeled as DAGs of primitives with an optional reinforcement module to incorporate prior knowledge. A GUI built on Axolotl/Orange enables drag-and-drop construction and evaluation, while a data-driven searcher automatically discovers high-performing pipelines, including data splitting and scoring steps, to yield deployment-ready solutions. The Apache 2.0–licensed system targets diverse detection tasks and aims to reduce human labor while accelerating real-world deployment.

Abstract

We present TODS, an automated Time Series Outlier Detection System for research and industrial applications. TODS is a highly modular system that supports easy pipeline construction. The basic building block of TODS is primitive, which is an implementation of a function with hyperparameters. TODS currently supports 70 primitives, including data processing, time series processing, feature analysis, detection algorithms, and a reinforcement module. Users can freely construct a pipeline using these primitives and perform end- to-end outlier detection with the constructed pipeline. TODS provides a Graphical User Interface (GUI), where users can flexibly design a pipeline with drag-and-drop. Moreover, a data-driven searcher is provided to automatically discover the most suitable pipelines given a dataset. TODS is released under Apache 2.0 license at https://github.com/datamllab/tods.

Figures (2)

• Figure 1: System overview. TODS provides end-to-end pipeline construction from data processing to detection algorithms. It also provides a reinforcement module to incorporate human knowledge into predictions.
• Figure 2: An illustration of the GUI. Users can easily build and evaluate a pipeline with drag-and-drop.