tmheo / ieee-fraud-detection-17th-place-solution Goto Github PK

Since the data is unbalanced and huge, I did under sampling right after reading the data.
Here is my code:
`isFraud_idx = y.loc[y==1].index
notFraud_idx = y.loc[y==0].index
n = len(isFraud_idx)
isFraud_idx = np.array(isFraud_idx)

rnd_notFraud_idx = np.random.choice(notFraud_idx, n, replace=False) # under sampling
rnd_notFraud_idx = np.array(rnd_notFraud_idx)
under_sample_idx = np.concatenate([rnd_notFraud_idx, isFraud_idx])

X = X.iloc[under_sample_idx,:]
y = y.iloc[under_sample_idx]`