CODEWITHSUNDEEP
pythonpandas
Krypt
Krypt

Reputation: 421

Drop Columns with more than 60 Percent of "empty" Values in Pandas

I have got a dataframe like this:

import pandas as pd
data = {
    'c1': ['Test1','Test2','NULL','Test3',' ','Test4','Test4','Test1',"Test3"],
    'c2': [' ','Test1',' ','NULL',' ','NULL','NULL','NULL','NULL'],
    'c3': [0,0,0,0,0,1,5,0,0],
    'c4': ['NULL', 'Test2', 'Test1','Test1', 'Test2', 'Test2','Test1','Test1','Test2']
}
df = pd.DataFrame(data)
df

The dataframe looks like this:

    c1      c2      c3      c4
0   Test1           0       NULL
1   Test2   Test1   0       Test2
2   NULL            0       Test1
3   Test3   NULL    0       Test1
4                   0       Test2
5   Test4   NULL    1       Test2
6   Test4   NULL    5       Test1
7   Test1   NULL    0       Test1
8   Test3   NULL    0       Test2

I want to drop all columns, that have more than 60 % of "empty" values. "Empty" means in my case that the values are for example: ' ', 'NULL' or 0. There are strings (c1, c2, c4) as well as integers (c3).

The result should be a dataframe with columns c1 and c4 only.

    c1      c4
0   Test1   NULL
1   Test2   Test2
2   NULL    Test1
3   Test3   Test1
4           Test2
5   Test4   Test2
6   Test4   Test1
7   Test1   Test1
8   Test3   Test2

I have no idea how to handle that problem. Only thing that comes to my mind is something like 

df.loc[:, (df != 0).any(axis=0)]

to delete all columns where all values are 0, 'NULL' and so on.

Upvotes: 19

Views: 42328

Answers (3)

Gurudev Kumar
Gurudev Kumar

Reputation: 199

Below given solution is very small and fast(in performance)

Step:1 we are finding percentage of null value in every column

Step:2 we are finding column names in list having more than 60% null values

Step:3 Drop columns having more than 60% null values

import pandas as pd

data = {
    'c1': ['Test1','Test2','NULL','Test3',' ','Test4','Test4','Test1',"Test3"],
    'c2': [' ','Test1',' ','NULL',' ','NULL','NULL','NULL','NULL'],
    'c3': [0,0,0,0,0,1,5,0,0],
    'c4': ['NULL', 'Test2', 'Test1','Test1', 'Test2', 'Test2','Test1','Test1','Test2']
}
df = pd.DataFrame(data)

# Below code gives percentage of null in every column
null_percentage = df.isnull().sum()/df.shape[0]*100

# Below code gives list of columns having more than 60% null
col_to_drop = null_percentage[null_percentage>60].keys()

output_df = df.drop(col_to_drop, axis=1)

Upvotes: 4

shivsn
shivsn

Reputation: 7838

you can drop the columns using dropna thresh parameter:

In [58]: df = df.replace([0,' ','NULL'],np.nan)
In[59]: df
Out[59]: 
      c1     c2   c3     c4
0  Test1    NaN  NaN    NaN
1  Test2  Test1  NaN  Test2
2    NaN    NaN  NaN  Test1
3  Test3    NaN  NaN  Test1
4    NaN    NaN  NaN  Test2
5  Test4    NaN  1.0  Test2
6  Test4    NaN  5.0  Test1
7  Test1    NaN  NaN  Test1
8  Test3    NaN  NaN  Test2

In [60]: df.dropna(thresh=df.shape[0]*0.6,how='all',axis=1)
Out[60]: 
      c1     c4
0  Test1    NaN
1  Test2  Test2
2    NaN  Test1
3  Test3  Test1
4    NaN  Test2
5  Test4  Test2
6  Test4  Test1
7  Test1  Test1
8  Test3  Test2

Upvotes: 21

jezrael
jezrael

Reputation: 863116

Use DataFrame.isin for check all formats and then get mean for treshold and filter by boolean indexing with loc:

print (df.isin([' ','NULL',0]))
      c1     c2     c3     c4
0  False   True   True   True
1  False  False   True  False
2   True   True   True  False
3  False   True   True  False
4   True   True   True  False
5  False   True  False  False
6  False   True  False  False
7  False   True   True  False
8  False   True   True  False

print (df.isin([' ','NULL',0]).mean())
c1    0.222222
c2    0.888889
c3    0.777778
c4    0.111111
dtype: float64

df = df.loc[:, df.isin([' ','NULL',0]).mean() < .6]
print (df)
      c1     c4
0  Test1   NULL
1  Test2  Test2
2   NULL  Test1
3  Test3  Test1
4         Test2
5  Test4  Test2
6  Test4  Test1
7  Test1  Test1
8  Test3  Test2

Upvotes: 25

Related Questions