Replies: 2 comments 1 reply
-
The way you set it up, you are asking Since there is a single machine, and you are asking It should work as you intend if you have 2 or more machines to cluster. Though I'm not sure if that's what you want - perhaps the task you are interested in is not time series clustering? Do you want to apply the clustering per time index, perhaps? That would be unsupervised annotation with categories, or segmentation. That module exists, but it is experimental. @Alex-JG3 is currently reworking interfaces, and some summer mentees are working on it. |
Beta Was this translation helpful? Give feedback.
-
@fkiraly update from my side on outlier detection in sktime PS. |
Beta Was this translation helpful? Give feedback.
-
I want to construct the following pipeline in sktime
I want it to behave exactly the same as my pipeline in sklearn
Currently, sktime doesn't work very well and sklearn works like a breeze but I can't include operations like detrending etc.
I provide the following toy dataset. I have Machines, which generate features on each time step.
Basically, I am interested in anomalies and want to cluster events in for example 3 groups. This to help the business in focusing on the right problems.
A set is as follows
Now in sktime
The first step goes fine
Hereafter it stops working. The pipeline does accept data when i do the following.
But it is not what I want
Another solution I see is
This works well until the PCA but fails at the kmeans as this no longer accepts 2 and raises the error.
Beta Was this translation helpful? Give feedback.
All reactions