CODEWITHSUNDEEP
pythonpandasdataframepandas-groupbybinning
NeStack
NeStack

Reputation: 2014

Pandas df: group, bin and average in different columns?

I have data that looks qualitatively like this dummy table:

speed_observation, car_brand, traction_force
10, ford, 2
20, ford, 4
35, seat, 8
50, ford, 16
10, audi, 2
20, audi, 5
43, audi, 2
12, seat, 2.5
10, ford, 0.5
30, audi, 6
23, ford, 4
17, seat, 5.5
10, seat, 10
38, audi, 2
40, ford, 9
19, ford, 6.6
49, seat, 18
18, ford, 4

I would like to group the dataframe by car brands and for each brand to bin the speed observations into ranges (e.g. [0,25] and [25,50]) and then for each brand and bin to calculate the average traction measured, receiving something like:

speed_bin_upper_lim, car_brand, avrg_traction_force_in_speed_bin
25, audi, X1
50, audi, X2
25, ford, X3
50, ford, X4
25, seat, X5
50, seat, X6

How do I do this? It should work for an arbitrary number of unique car_brand classes and the user should only provide the number of speed bins or the range of the bins (e.g. n=3 or [0,25,50]). I suppose pd.groupby and pd.cut will do it, but I didn't find how exactly.

Quang Hoang's answer works very well, and if you want to extend it, because you want to group by one more column, let's say wheel_kind, and your dataframe looks like:

speed_observation,car_brand,wheel_kind,traction_force
10, ford, winter, 2
20, ford, summer, 4
35, seat, summer, 8
50, ford, winter, 16
10, audi, summer, 2
20, audi, summer, 5
43, audi, summer, 2
12, seat, summer, 2.5
10, ford, summer, 0.5
30, audi, summer, 6
23, ford, summer, 4
17, seat, summer, 5.5
10, seat, summer, 10
38, audi, summer, 2
40, ford, summer, 9
19, ford, summer, 6.6
49, seat, summer, 18
18, ford, summer, 4

Then just add the column wheel_kind into the previous solution, more precisely:

(df.groupby(['car_brand', `wheel_kind`, cuts])
   .traction_force.mean()
   .reset_index(name='avg_traction_force')
)

and afterwards don't forget to drop the NaNs, because for ford and audi there are no winter wheels:

df_grp.dropna(inplace=True)
df_grp.reset_index(drop=True, inplace=True) #just to reset the index

Upvotes: 0

Views: 1941

Answers (2)

ansev
ansev

Reputation: 30920

We can create a series to group manually as an alternative to pd.cut

n = 25

blocks = (df.speed_observation.sub(1) // n).add(1).mul(n)
blocks = blocks.rename('speed_bin_upper_lim') 

(df.groupby([blocks, 'car_brand'])
   .traction_force.mean()
   .reset_index(name='avrg_traction_force_in_speed_bin'))

   speed_bin_upper_lim car_brand  avrg_traction_force_in_speed_bin
0                   25      audi                          3.500000
1                   25      ford                          3.516667
2                   25      seat                          6.000000
3                   50      audi                          3.333333
4                   50      ford                         12.500000
5                   50      seat                         13.000000

Detail

print(blocks)
0     25
1     25
2     50
3     50
4     25
5     25
6     50
7     25
8     25
9     50
10    25
11    25
12    25
13    50
14    50
15    25
16    50
17    25
Name: speed_bin_upper_lim, dtype: int64

Upvotes: 1

Quang Hoang
Quang Hoang

Reputation: 150735

You can just cut the speed_observation with the desired bins and group by that:

cuts = pd.cut(df['speed_observation'], [0,25,50])

(df.groupby(['car_brand', cuts])
   .traction_force.mean()
   .reset_index(name='avg_traction_force')
)

Output:

  car_brand speed_observation  avg_traction_force
0      audi           (0, 25]            3.500000
1      audi          (25, 50]            3.333333
2      ford           (0, 25]            3.516667
3      ford          (25, 50]           12.500000
4      seat           (0, 25]            6.000000
5      seat          (25, 50]           13.000000

Upvotes: 2

Related Questions