Feedback on a new dataset

[rajatanand86]

Hi
I have been able to successfully reproduce the accuracy of CHBMIT patient 20 dataset.
Now, I am trying this algorithm on a new dataset. But able to get 0.45 area under roc curve.
Some details on dataset.
ictal_X[0].shape=(126, 1, 22, 55, 114)
len(ictal_X)=9
interictal_X[0].shape=(1543, 1, 22, 55, 114)
len(interictal_X)=6
In the sampling_CHBMIT.csv, I have kept ictal_ovl=1 and intericatl_unspl=1 to get above stats. (need your feedback on this as I dont know how to set this up)

Would be great to get guidance on how to improve roc.