CODEWITHSUNDEEP
pythonpandaspandas-groupby
Outcast
Outcast

Reputation: 5117

Change value in function used by pandas groupby

I am doing the following:

def percentage(x):
    return x[(x<=5)].count() / x.count() * 100

full_data = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage})

But I want to do this groupby successively with multiple values such x<=7, x<=9, x<=11 etc at the percentage function.

What is the easiest way to do this instead of writing multiple functions and calling them?

So basically I want to avoid doing something like this:

def percentage_1(x):
    return x[(x<=5)].count() / x.count() * 100

full_data_1 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage_1})

def percentage_2(x):
    return x[(x<=7)].count() / x.count() * 100

full_data_2 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage_2})

# etc.

Upvotes: 2

Views: 100

Answers (2)

Outcast
Outcast

Reputation: 5117

I came up with this as the most concise solution to my question:

def percentage(x):
    global c
    return x[(x<=c)].count() / x.count() * 100

c=5
full_data_5 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage})

c=7
full_data_7 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage})

c=9
full_data_9 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage})

# etc

However, I am using a global variable and this is a quite controversial practice.

Upvotes: 0

jezrael
jezrael

Reputation: 863301

You can rewrite your function - create new column filled by boolean mask and then aggregate mean and last multiple by 100 with Series.mul:

n = 3

full_data['new'] = full_data['Volume'] <= n
full_data = full_data.groupby(['Id', 'Week_id'])['new'].mean().mul(100).reset_index()

Solution with function:

def per(df, n):
    df['new'] = df['Volume'] <= n
    return df.groupby(['Id', 'Week_id'])['new'].mean().mul(100).reset_index()

EDIT: Solution from github:

full_data = pd.DataFrame({
        'Id':list('XXYYZZXYZX'),
         'Volume':[2,4,8,1,2,5,8,2,6,4],
         'Week_id':list('aaabbbabac')
})

print (full_data)

val = 5
def per(c):
    def f1(x):
        return x[(x<=c)].count() / x.count() * 100
    return f1

full_data2 = full_data.groupby(['Id', 'Week_id']).agg({'Volume': per(val)}).reset_index()
print (full_data2)
  Id Week_id      Volume
0  X       a   66.666667
1  X       c  100.000000
2  Y       a    0.000000
3  Y       b  100.000000
4  Z       a    0.000000
5  Z       b  100.000000

def percentage(x):
    return x[(x<=val)].count() / x.count() * 100

full_data1 = full_data.groupby(['Id', 'Week_id'], as_index=False).agg({'Volume': percentage})

print (full_data1)
  Id Week_id      Volume
0  X       a   66.666667
1  X       c  100.000000
2  Y       a    0.000000
3  Y       b  100.000000
4  Z       a    0.000000
5  Z       b  100.000000

Upvotes: 2

Related Questions