That's a good question. `df_feats_validation` contains exactly the same features as `df_feats`, only for points in validation time interval. Yes I probably could put the whole code together to give better understanding. There won't be much new code though, as most of it is already included in the article, just fragmented.