So, I need to create a classifier with 3 simple comparisons to detect a fake bill, based on something like this pseudocode:
assume you are examining a bill with features f_1 ,f_2 ,f_3 and f_4 your rule may look like this :
if ( f_1 > 4) and ( f_2 > 8) and ( f_4 < 25):
x = " good "
else :
x = " fake "
What is best to use for this - a lambda? I started with this:
distdf = {
f1 : banknote['variance']
f2 : banknote['skewness']
f3 : banknote['curtosis']
f4 : banknote['entropy']
}
But I am not sure how to proceed. This is using the famous bank note authentication dataset: BankNote_Authentication.csv that can be found on Kaggle.
We can try with np.where
to check all conditions and apply the corresponding labels. No need to alias columns to f1
, f2
, f3
etc:
banknote_df['classifier'] = np.where(
(banknote_df['variance'] > 4) &
(banknote_df['skewness'] > 8) &
(banknote_df['entropy'] < 25),
'good',
'fake'
)
Sample Program:
import numpy as np
import pandas as pd
banknote_df = pd.DataFrame({
'variance': [2.2156, 4.4795, 1.866, 3.47578, 0.697854],
'skewness': [9.45647, 8.54688, -5.4568, 6.15258, -3.4564],
'curtosis': [-1.12245, -1.2454, 2.75, -6.5468, 3.45875],
'entropy': [-0.424514, -2.45687, 0.1230152, -6.1254, -0.45241],
'class': [0, 0, 0, 0, 0]
})
banknote_df['classifier'] = np.where(
(banknote_df['variance'] > 4) &
(banknote_df['skewness'] > 8) &
(banknote_df['entropy'] < 25),
'good',
'fake'
)
print(banknote_df)
banknote_df
:
variance skewness curtosis entropy class classifier
0 2.215600 9.45647 -1.12245 -0.424514 0 fake
1 4.479500 8.54688 -1.24540 -2.456870 0 good
2 1.866000 -5.45680 2.75000 0.123015 0 fake
3 3.475780 6.15258 -6.54680 -6.125400 0 fake
4 0.697854 -3.45640 3.45875 -0.452410 0 fake