About anomaly detection

[PoleGeogry]

I would like to ask, UCR data is multiple univariate time series data. If I want to use multi-variable time series anomaly detection, how should I modify the running script or how to modify my custom dataset?

[ANopony]

There is an argument "feature", maybe we can change it to use multi-variable time series.

[PoleGeogry]

@ANopony Sorry, I couldn't achieve my effect by using feature parameters. I would like to ask how to use public multi-variable time series data sets such as SWAT, PSM, etc. for anomaly detection, using Timer.

[ANopony]

Actually I used a multi-variable time series data set for anomaly detection a two weeks ago, and it trained successfully. My dataset has one timestamp column and three feature columns named AccX, AccY and AccZ. Perhaps code cannot get the border to divide SWAT, PSM, etc. Would you mind paste the error the terminal said?

[PoleGeogry]

@ANopony Sorry to bother you, could you please take a look at your running script?

[ANopony]

Sure，feel free.
Do you mean this one?

[PoleGeogry]

@ANopony Thank you for sharing. Since I have reproduced Time-Series-Library before, I thought that the multi-variable csv file can be directly passed in as a dataset. This has been bothering me for a long time, thanks for your guidance.