CODEWITHSUNDEEP
pythonpandasdataframe
jtitusj
jtitusj

Reputation: 3084

Using pandas apply

I have a dataframe that is similar to the example below:

sample = {'col1': [50.6, 30.67, 40.5, 0, 0, 0],
          'col2': [40.74, 30.33, 41.00, 0, 0, 0]}
df_sample = pd.DataFrame(sample)

Now, in both col2 and col3 however, the entries represent two different values. For example, for entry 50.6, that represents val1 = 5 and val2 = 0.6. Another example would be 41.00. This value represents 4 and 1.0.

Basically, what I want to get is a column which can be computed as follows:

df_sample['res'] = df_sample.apply(lambda x: 
    ((x['col2']//10)*(x['col2']%10) + (x['col3']//10)*(x['col3']%10)) 
            / (x['col2']//10 + x['col3']//10), axis=1)
df_sample.fillna(0)

Basically, it gets the weighted average from the values obtained from each column. Now, what I want to do is scale this method to work with let's say twenty columns without hardcoding each column name in the DataFrame. Please advise.

Upvotes: 1

Views: 50

Answers (2)

jezrael
jezrael

Reputation: 863531

You can omit apply and rather use Series (columns of Dataframes):

sample = {'col2': [50.6, 30.67, 40.5, 0, 0, 0],
          'col3': [40.74, 30.33, 41.00, 0, 0, 0],
          'col4': [70.6, 80.67, 70.5, 0, 0, 0],
          'col5': [10.74, 50.33, 51.00, 0, 0, 0]}
df_sample = pd.DataFrame(sample)
print (df_sample)
    col2   col3   col4   col5
0  50.60  40.74  70.60  10.74
1  30.67  30.33  80.67  50.33
2  40.50  41.00  70.50  51.00
3   0.00   0.00   0.00   0.00
4   0.00   0.00   0.00   0.00
5   0.00   0.00   0.00   0.00

I think you need:

print ((((df_sample['col2']//10 * df_sample['col2']%10) + 
        (df_sample['col3']//10 * df_sample['col3']%10) +
        (df_sample['col4']//10 * df_sample['col4']%10) +
        (df_sample['col5']//10 * df_sample['col5']%10)) 
         / (df_sample['col2']//10 + df_sample['col3']//10 + 
            df_sample['col4']//10 + df_sample['col5']//10)).fillna(0))

0    0.641176
1    0.526842
2    0.725000
3    0.000000
4    0.000000
5    0.000000
dtype: float64

print (((df_sample//10 * df_sample%10).sum(axis=1).div((df_sample//10).sum(axis=1)))
         .fillna(0))
0    0.641176
1    0.526842
2    0.725000
3    0.000000
4    0.000000
5    0.000000
dtype: float64

Timings:

In [114]: %timeit ((((df_sample['col2']//10 * df_sample['col2']%10) + (df_sample['col3']//10 * df_sample['col3']%10) + (df_sample['col4']//10 * df_sample['col4']%10) + (df_sample['col5']//10 * df_sample['col5']%10))  / (df_sample['col2']//10 + df_sample['col3']//10 + df_sample['col4']//10 + df_sample['col5']//10)).fillna(0))
100 loops, best of 3: 2.03 ms per loop

In [115]: %timeit (((df_sample//10 * df_sample%10).sum(axis=1).div((df_sample//10).sum(axis=1))).fillna(0))
1000 loops, best of 3: 897 µs per loop

Upvotes: 0

Nickil Maveli
Nickil Maveli

Reputation: 29719

Just create a subset of the columns you want to use for computing and you can perform the operation on the subsetted df itself and not calling functions on every series object:

np.random.seed(42)
df = pd.DataFrame(np.random.uniform(0, 100, (100, 25))).add_prefix('col')
df.shape
(100, 25)

# Take first 20 columns (for eg)
df_sample = df.iloc[:, :20] 
df['res'] = (df_sample // 10 * df_sample % 10).sum(1)/(df_sample // 10).sum(1)

Upvotes: 1

Related Questions