I am trying to understand the TimeSeriesKMeans. My idea is to obtain a non-supervised model with traditional features and time series features. I would like to use TimeSeriesKMeans. I know there is a library that transforms time series into classical features not affected by time (mean, min, max...) and you can use traditional unsupervised methods but I would like to use DWT. I was wondering if I can combine normal features with time series features?, would it be repeating the same feature for the same identifier?. Is that efficient?.
Thank you