Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RAMSeS: Robust and Adaptive Model Selection for Time-Series Anomaly Detection Algorithms

Mohamed Abdelmaksoud, Sheng Ding, Andrey Morozov, Ziawasch Abedjan

TL;DR

This work presents the Robust and Adaptive Model Selection for Time-Series Anomaly Detection RAMSeS framework, and evaluates RAMSeS and shows that it outperforms prior methods on F1.

Abstract

Time-series data vary widely across domains, making a universal anomaly detector impractical. Methods that perform well on one dataset often fail to transfer because what counts as an anomaly is context dependent. The key challenge is to design a method that performs well in specific contexts while remaining adaptable across domains with varying data complexities. We present the Robust and Adaptive Model Selection for Time-Series Anomaly Detection RAMSeS framework. RAMSeS comprises two branches: (i) a stacking ensemble optimized with a genetic algorithm to leverage complementary detectors. (ii) An adaptive model-selection branch identifies the best single detector using techniques including Thompson sampling, robustness testing with generative adversarial networks, and Monte Carlo simulations. This dual strategy exploits the collective strength of multiple models and adapts to dataset-specific characteristics. We evaluate RAMSeS and show that it outperforms prior methods on F1.

RAMSeS: Robust and Adaptive Model Selection for Time-Series Anomaly Detection Algorithms

TL;DR

This work presents the Robust and Adaptive Model Selection for Time-Series Anomaly Detection RAMSeS framework, and evaluates RAMSeS and shows that it outperforms prior methods on F1.

Abstract

Time-series data vary widely across domains, making a universal anomaly detector impractical. Methods that perform well on one dataset often fail to transfer because what counts as an anomaly is context dependent. The key challenge is to design a method that performs well in specific contexts while remaining adaptable across domains with varying data complexities. We present the Robust and Adaptive Model Selection for Time-Series Anomaly Detection RAMSeS framework. RAMSeS comprises two branches: (i) a stacking ensemble optimized with a genetic algorithm to leverage complementary detectors. (ii) An adaptive model-selection branch identifies the best single detector using techniques including Thompson sampling, robustness testing with generative adversarial networks, and Monte Carlo simulations. This dual strategy exploits the collective strength of multiple models and adapts to dataset-specific characteristics. We evaluate RAMSeS and show that it outperforms prior methods on F1.
Paper Structure (45 sections, 10 equations, 10 figures, 10 tables, 3 algorithms)

This paper contains 45 sections, 10 equations, 10 figures, 10 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Time Series Anomaly Detection and Model Selection
  4. Surrogate Metrics for Model Selection
  5. Learning-Based Adaptive Selection
  6. Meta-Learning for Time Series Model Choice
  7. Ensemble Selection
  8. Benchmarking-Based Evaluation Paradigms
  9. Problem Formulation and RAMSeS Overview
  10. Problem formulation
  11. RAMSeS Overview
  12. Ensemble Branch ❶ in Fig. \ref{['fig:Architecture']}
  13. Single-Model Selection Branch ❷ in Fig. \ref{['fig:Architecture']}
  14. LinTS in Fig. \ref{['fig:Architecture']}
  15. Robustness-based tests in Fig. \ref{['fig:Architecture']}
  16. ...and 30 more sections

Figures (10)

  • Figure 1: F1 score for six anomaly detection models from statistical and neural-network families across four domains (Prices, Medical, Industry, and Environment) ucrskabsmdvandenburg2020evaluation_TCPD. Red vertical lines indicate spike anomalies. The triangle marks the best-performing model in each domain.
  • Figure 2: RAMSeS framework overview.
  • Figure 3: SKAB 1-1 time series
  • Figure 4: Stacking ensemble via a genetic algorithm.
  • Figure 5: Linear Thompson sampling with $\varepsilon$-greedy for model selection.
  • ...and 5 more figures