Python count and probability

Question

I have the following data :

Name    Item
peter   apple
peter   apple
Ben     banana
peter   banana

I want to print

frequency of what peter eat :
apple 2 
banana 1 

this is my code

u, count = np.unique(data['Item'], return_counts=True)

process = u[np.where(data['Name']= 'peter')[0]]

process2 = dict(Counter(process))
print "Item\frequency"

for k, v in process2.items():
print '{0:.0f}\t{1}'.format(k,v)

but it got error I also want to calculate the probability of peter eat apple next time but I dont have any idea , any suggestion ?

Answer 1

The error you are getting is as the other answer indicates, you cannot use data['Name'] = 'peter' as a function parameter, you actually intended to use - np.where(data['Name'] == 'peter') .

But, given that you are using pandas , and I am guessing data is a pandas DataFrame . In which case, what you really want can be achieved using DataFrame.groupby. Example -

data[data['Name']=='peter'].groupby('Item').count()

Demo -

In [7]: data[data['Name']=='peter'].groupby('Item').count()
Out[7]:
        Name
Item
apple      2
banana     1

---

If you want this printed in a loop, you can use -

df = data[data['Name']=='peter'].groupby('Item').count()
for fruit,count in df['Name'].iteritems():
    print('{0}\t{1}'.format(fruit,count))

Demo -

In [24]: df = data[data['Name']=='peter'].groupby('Item').count()

In [25]: for fruit,count in df['Name'].iteritems():
   ....:     print('{0}\t{1}'.format(fruit,count))
   ....:
apple   2
banana  1

---

For the updated issue that the OP was getting, where he was getting the following error -

TypeError: invalid type comparison

The issue occurs in this case because in the real data for the OP , the column has numeric values (float/int) , but the OP was comparing the values against string, and hence getting the error. Example -

In [30]: df
Out[30]:
   0  1
0  1  2

In [31]: df[0]=='asd'
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-31-e7bacd79d320> in <module>()
----> 1 df[0]=='asd'

C:\Anaconda3\lib\site-packages\pandas\core\ops.py in wrapper(self, other, axis)
    612
    613             # scalars
--> 614             res = na_op(values, other)
    615             if np.isscalar(res):
    616                 raise TypeError('Could not compare %s type with Series'

C:\Anaconda3\lib\site-packages\pandas\core\ops.py in na_op(x, y)
    566                 result = getattr(x, name)(y)
    567                 if result is NotImplemented:
--> 568                     raise TypeError("invalid type comparison")
    569             except (AttributeError):
    570                 result = op(x, y)

TypeError: invalid type comparison

If your column is numeric, you should compare against numeric values, not string.

Answer 2

You can groupby the name and use value_counts:

In [11]: df.groupby("Name")["Item"].value_counts()
Out[11]:
Name
Ben    banana    1
peter  apple     2
       banana    1
dtype: int64

Potentially you could unstack these into columns:

In [12]: df.groupby("Name")["Item"].value_counts().unstack(1)
Out[12]:
       apple  banana
Name
Ben      NaN       1
peter      2       1

In [13]: res = df.groupby("Name")["Item"].value_counts().unstack(1).fillna(0)

In [13]: res
Out[13]:
       apple  banana
Name
Ben        0       1
peter      2       1

To get the probabilities divide by the sum:

In [14]: res = res.div(res.sum(axis=1), axis=0)

In [15]: res
Out[15]:
          apple    banana
Name
Ben    0.000000  1.000000
peter  0.666667  0.333333

and the probability peter eats an apple next time:

In [16]: res.loc["peter", "apple"]
Out[16]: 0.66666666666666663