I have a sample dataset with about a million records with about 7 feature variables. On running logistic regression on this dataset, I got 5 features with a positive coefficient and 2 with a negative coefficient.
Now I need to extrapolate this data to about 50 times and add a couple of feature variables to the set. How can I ensure that after generating random data and adding more features, I achieve maximum features with positive coefficients?
Is there a way to identify patterns or relationships that I could use to define rules for replicating the data in the way that I want ?
If this sounds a bit vague, I'd be happy to have a detailed discussion on this if possible.