some statistics
1 view
Skip to first unread message
Joseph Picone
May 1, 2024, 10:25:09 AMMay 1
to ECE 8527
Some statistics on how frequently each condition occurs in the TNMG
database:
1dAVb, RBBB, LBBB, SB, AF, ST
0: 2,286707 2,258,940 2,285,298 2,284,517 2,280,614 2,272,593
1: 35,758 63,525 37,167 37,948 41,851 49,872
If you train on only unhealthy patients, the imbalance in the data
should not be a huge issue.
The overall data was balanced so that there were equal numbers of
unhealthy and healthy patients. So the healthy class ("000000") occurs 6
times greater than the unhealthy classes when you pool the training data.
I don't think that will cause problems, but if it does, reduce the
number of healthy patients when you train.
-Joe
database:
1dAVb, RBBB, LBBB, SB, AF, ST
0: 2,286707 2,258,940 2,285,298 2,284,517 2,280,614 2,272,593
1: 35,758 63,525 37,167 37,948 41,851 49,872
If you train on only unhealthy patients, the imbalance in the data
should not be a huge issue.
The overall data was balanced so that there were equal numbers of
unhealthy and healthy patients. So the healthy class ("000000") occurs 6
times greater than the unhealthy classes when you pool the training data.
I don't think that will cause problems, but if it does, reduce the
number of healthy patients when you train.
-Joe
Reply all
Reply to author
Forward
0 new messages