Pandas groupby get value of a column where another column is minimum

user_8876025 0
Viewed 562

I have a dataframe df like this:

region model metrics
Tokyo ARIMA 0.1
Tokyo FeedForward 0.2
Tokyo DeepAR 0.3
Osaka ARIMA 0.5
Osaka FeedForward 0.2
Osaka DeepAR 0.1

I want to group this by region and return the minimum value of metrics in each group, as well as the model value where the metrics is minimum.

The expected result:

| region | model | metrics|
| -------- | --------- |----|
| Tokyo    | ARIMA |0.1|
| Osaka    | DeepAR    |0.1|

I tried to do it like below, but not sure how I can complete:

df.groupby("region").agg({'metrics':'min', ####... })

Maybe use argmin? Any help will be appreciated. Thanks!

pandas group-by python
3 Answers

You can find the index of the minimum metric of each group and then loc with it into the original frame:

# not sorting to keep the original order of appearance of regions
min_inds = df.groupby("region", sort=False).metrics.idxmin()
result = df.loc[min_inds]

to get

>>> result

  region   model  metrics
0  Tokyo   ARIMA      0.1
5  Osaka  DeepAR      0.1

(may the forecasting be fun!)

user_93321870

How about sort by value of metrics and drop duplicates remaining the smallest one like this.

df.sort_values("metrics").drop_duplicates(['region'], keep='first')
user_123699590 
df.groupby("region").agg({'metrics':['min']})
user_126986060
Related