How can I use value_counts() only for certain values?-CodePudding

I want to extract how many positive reviews by brand are in a dataset which includes reviews from thousands of products. I used this code and I got a table including percentaje of positive and non-positive reviews. How can I get only the percentage of positive reviews by brand? I only want the "True" results in positive_review. Thanks!

df_reviews_ok.groupby("brand")["positive_review"].value_counts(normalize=True).mul(100).round(2)
brand                   positive_review
Belkin                  False               70.00
                        True                30.00
Bowers & Wilkins        False               67.65
                        True                32.35
Corsair                 False               75.22
                        True                24.78
Definitive Technology   False               68.29
                        True                31.71
Dell                    False               60.87
                        True                39.13
DreamWave               False              100.00
House of Marley         False              100.00
JBL                     False               58.43
                        True                41.57
Kicker                  True                66.67
                        False               33.33
Lenovo                  False               76.92
                        True                23.08
Logitech                False               75.75
                        True                24.25
MEE audio               False               53.80
                        True                46.20
Microsoft               False               67.86
                        True                32.14
Midland                 False               72.09
                        True                27.91
Motorola                False               72.92
                        True                27.08
Netgear                 False               72.30
                        True                27.70
Pny                     False               68.78
                        True                31.22
Power Acoustik          False              100.00
SVS                     False              100.00
Samsung                 False               61.94
                        True                38.06
Sanus                   False               75.93
                        True                24.07
Sdi Technologies, Inc.  False               55.63
                        True                44.37
Siriusxm                False               73.33
                        True                26.67
Sling Media             False               67.16
                        True                32.84
Sony                    False               55.40
                        True                44.60
Toshiba                 False               56.52
                        True                43.48
Ultimate Ears           False               70.21
                        True                29.79
Verizon Wireless        False               75.86
                        True                24.14
WD                      False               58.33
                        True                41.67
Yamaha                  False               61.15
                        True                38.85
Name: positive_review, dtype: float64

CodePudding user response：

Using the following toy DataFrame as an example:

df = pd.DataFrame({
    'brand': list('AAAABBBB'),
    'positive': [True, True, False, False, True, True, True, False]
})

If you would like to get the percentage of positive reviews for each brand relative to the total number of reviews per brand then try:

df.groupby('brand')['positive'].mean()

The result is as expected:

brand
A    0.50
B    0.75
Name: positive, dtype: float64

CodePudding user response：

You can unstack the output and slice the True

(df.groupby('brand')
   ['positive_review'].value_counts(normalize=True)
   .mul(100).round(2)
   .unstack(fill_value=0)
   [True]
 )

CodePudding user response：

How about using .reset_index() after the statement and then using a condition.