CODEWITHSUNDEEP
pythonpython-3.xpandas
sab
sab

Reputation: 97

Select row with MAX value if conditions are met

I am using Pandas in Python and having a spot of trouble. I have a dataframe with an index and 2 columns: "VIFFactor" and "features".

I am trying to return the "feature" record from my dataframe that has the maximum value in "VIFFactor" PROVIDED the number is over 5.

Here's my code:

I have tried to replace idsmax with max and got "cannot compare a dtyped [bool] array with a scalar of type [bool]".

vif3 = vif_test.loc[(vif_test['VIFFactor'] >= 5) & (vif_test['VIFFactor'].idxmax()), 'features']

So for example, from the below, I would like to return HadCampaign because it is the highest record and over 5 but I am currently getting nothing:


    VIFFactor   features
2   12.028754355028974  HadCampaign
22  11.98926492333954   DiscountedPrice
29  5.460195615389739   RatingsReceivedRank
30  4.59851607313422    SortOrder
19  3.0681452496804833  PreferredPartnerBadge
9   3.0554578279939815  PerkCustomerDropService
28  2.735597253984768   RatingsReceived
26  2.263922204962396   PriceRank

Upvotes: 0

Views: 4924

Answers (2)

BENY
BENY

Reputation: 323226

This will not work you should look at max , your first condition pass a bool and the length is len(df), but idxmax will only return the index meet the max value which is one value , so if you need slice the dataframe , you need pass two condition same length at least 

 vif3 = vif_test.loc[(vif_test['VIFFactor'] >= 5) & (vif_test['VIFFactor'].max() == vif_test['VIFFactor']), 'features']

Upvotes: 1

BarefootDev
BarefootDev

Reputation: 346

Splitting over two lines might be clearer:

vif_test = vif_test[vif_test.VIFFactor > 5].set_index('VIFFactor')
vif3 = vif_test.loc[max(vif_test.index), 'feature']

Note: max() may often be faster that index.max(). Testing on a small dataframe:

%timeit d.index.max()
34.3 µs ± 447 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

 %timeit max(d.index)
9.43 µs ± 143 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)

Upvotes: 0

Related Questions