countzro's comments

On: Time-Series Anomaly Detection: A Decade Review

[countzro]

I also liked the main idea of Series2Graph but found the implementation a bit complicated.

There is a similar algorithm with a simpler implementation in this paper: „GraphTS: Graph-represented time series for subsequence anomaly detection“ https://pmc.ncbi.nlm.nih.gov/articles/PMC10431630/

The approach is for univariate time series and I found it to perform well (with very minor tweaks).

[mathewshen → countzro]

Agree. And thanks for the recommendation, I'd like to read the paper of GraphTS.