Count some occurrences grouping data with Pandas

user_8193292 0
Viewed 52

I've a Dataset structured like this: 

id    date       body        sentiment
1     1/1/2018   Some Text   Positive
2     1/1/2018   Some Text   Negative
3     1/1/2018   Some Text   None
4     1/2/2018   Some Text   Positive
5     1/2/2018   Some Text   None

For each day, I've some information like body (general text) and an associated sentiment. I would like to know, for each day, the count of positive, negative and None (without values) rows.

I've tried something like this but doesn't work: 

df.groupby('date', 'sentiment').count()

df['positive'] = df.groupby('date', 'sentiment').apply(lambda x: x is  Positive).count()

Any ideas?

pandas grouping python
1 Answers

Use crosstab with merge and left join:

df = df.merge(pd.crosstab(df['date'], df['sentiment']), on='date', how='left')
print (df)
   id      date       body sentiment  Negative  None  Positive
0   1  1/1/2018  Some Text  Positive         1     1         1
1   2  1/1/2018  Some Text  Negative         1     1         1
2   3  1/1/2018  Some Text      None         1     1         1
3   4  1/2/2018  Some Text  Positive         0     1         1
4   5  1/2/2018  Some Text      None         0     1         1

Detail:

print (pd.crosstab(df['date'], df['sentiment']))
sentiment  Negative  None  Positive
date                               
1/1/2018          1     1         1
1/2/2018          0     1         1

Another solution with GroupBy.size and join:

df = df.join(df.groupby(['date', 'sentiment']).size().unstack(fill_value=0), on='date')
user_29010020
Related