Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
I have recently googled the best proportion for training and test set for classifying physiological data between normal and abnormal. Much of the source tells that the proportion should be 70:30 or 80:20 for train:test. One of the sources could be visited here.
I am a little unsure because I haven't found the ground scientific truth behind this proportion. Is that any science that explains this proportion?
Thanks
In short theres no best proportion or rule of thumb.
It's highly dependent on how much data you have, its distribution in relation to the number of labels, and if samples are related to each other in anyway or are they completely independent of each other. You could look into k-fold training. Say you split your data into 5 sections each 20 percent of the size of the entire dataset. For each split of 20 percent of data use the remaining 80 percent of data as the training set. Go through each 20 percent of data and use that as the test and the rest as the training. Make sure your labels are evenly distributed between splits.
