CODEWITHSUNDEEP
pythonpython-3.xdataframeoutliers
Rahul Sharma
Rahul Sharma

Reputation: 2495

Remove outliers from a certain column

I have a Dataframe by the name bids_data

bids_data:

  Supplier_ID  shiper_RFQ
----------
0    2305      5000
1    2309      5200
2    2305      6500 
3    2307      4500
4    2301      900
5    2302      10000
6    2306      4500

and I want to remove the outliers rows from shiper_RFQ and store them in another dataframe. I tried converting the shiper_RFQ in a list and then finding the outliers but it doesn't work well.

Upvotes: 2

Views: 809

Answers (2)

Nihal
Nihal

Reputation: 5324

if you have good data then use threshold = 0.5

threshold = 1
print(df[df['shiper_RFQ'].apply(lambda x: np.abs(x - df['shiper_RFQ'].mean()) / df['shiper_RFQ'].std() < threshold)])

also this

 df = df[ np.abs(df['shiper_RFQ'] - df['shiper_RFQ'].mean()) / df['shiper_RFQ'].std() < threshold]

both will have same result

output

   Supplier_ID  shiper_RFQ
0         2305        5000
1         2309        5200
2         2305        6500
3         2307        4500
6         2306        4500

if you print you can see the anomaly 

print(df['shiper_RFQ'].apply(lambda x: np.abs(x - df['shiper_RFQ'].mean()) / df['shiper_RFQ'].std()))

0    0.084182
1    0.010523
2    0.468261
3    0.268329
4    1.594192
5    1.757294
6    0.268329

Upvotes: 1

Nathaniel
Nathaniel

Reputation: 3290

You can identify outliers by finding rows that differ from the mean column value by more than 1.5 standard deviations (or any other cut-off value you choose):

df['outliers'] = 0
df.loc[(df.shiper_RFQ - df.shiper_RFQ.mean()).abs() > 1.5*df.shiper_RFQ.std(), 'outliers'] = 1
print(df)

   Supplier_ID  shiper_RFQ  outliers
0         2305        5000         0
1         2309        5200         0
2         2305        6500         0
3         2307        4500         0
4         2301         900         1
5         2302       10000         1
6         2306        4500         0

Then you can store them in another data frame and remove them from the original:

df2 = df[df.outliers == 1].reset_index(drop=True)
df = df[df.outliers == 0].reset_index(drop=True)
print(df2)
print(df)

   Supplier_ID  shiper_RFQ  outliers
0         2301         900         1
1         2302       10000         1

   Supplier_ID  shiper_RFQ  outliers
0         2305        5000         0
1         2309        5200         0
2         2305        6500         0
3         2307        4500         0
4         2306        4500         0

Upvotes: 1

Related Questions