On time series clustering with k-means