Fun-TSG: A Function-Driven Multivariate Time Series Generator with Variable-Level Anomaly Labeling

📰 ArXiv cs.AI

Learn how to generate multivariate time series data with variable-level anomaly labeling using Fun-TSG, a function-driven approach, to improve anomaly detection methods

advanced Published 17 Apr 2026

Action Steps

Implement Fun-TSG to generate multivariate time series data with variable-level anomaly labeling
Use the generated data to train and test anomaly detection models
Evaluate the performance of different anomaly detection methods using the generated data
Compare the results to existing benchmark datasets
Apply Fun-TSG to real-world datasets to improve anomaly detection capabilities

Who Needs to Know This

Data scientists and machine learning engineers working on anomaly detection tasks can benefit from this research to evaluate and compare their models more effectively

Key Insight

💡 Fun-TSG provides a function-driven approach to generate multivariate time series data with fine-grained anomaly annotations, enabling more effective evaluation and comparison of anomaly detection models

Key Takeaways

Learn how to generate multivariate time series data with variable-level anomaly labeling using Fun-TSG, a function-driven approach, to improve anomaly detection methods

Full Article

Title: Fun-TSG: A Function-Driven Multivariate Time Series Generator with Variable-Level Anomaly Labeling

Abstract:
arXiv:2604.14221v1 Announce Type: new Abstract: Reliable evaluation of anomaly detection methods in multivariate time series remains an open challenge, largely due to the limitations of existing benchmark datasets. Current resources often lack fine-grained anomaly annotations, do not provide explicit intervariable and temporal dependencies, and offer little insight into the underlying generative mechanisms. These shortcomings hinder the development and rigorous comparison of detection models, es

Read full paper → ← Back to Reads