Assigning computed feature values to all rows belonging to a group

[kar10tik]

Tsfresh provides a lot of useful features from an input DataFrame (df). But I'm looking for something similar to pd.df.groupby.transform, which allows me to compute features for each group and helps assign, in a new column, the value of the computed feature to all rows belonging to that group:

groups = df.groupby(group_cols)
df["Std"] = groups["column"].transform('std') # Standard deviation as example

Pandas provides this groupby transform feature for a few statistical features. I wish to use tsfresh to compute comprehensive features and assign the feature values to each row belonging to the group in a new column. How can I do this?

[nils-braun]

You might be able to do this with pandas-own operations. I think using pd.groupby.apply with a extract_features call might be helpful here.

tmp = df.groupby(...).apply(lambda x: extract_features(x, ...).droplevel(0)

After that, the tmp dataframe has the "id" column as index, so you can just join it with the original data. I think this should give you what you want (?)