I am having the transaction dataset which contains 95% or user initiated transactions and 5% fraud transactions. When I am fitting the logistic regression model, it gives me bias prediction - as most of the transactions are as user initiated. So, I want to fix this issue of data distribution or other issue?
One of source suggested that prior probability can help in this issue. If this is true can any one explain? What's it and if this is not helpful then which are the possible solution for this problem?
I have other user profile variables but I want to get this thing solved without considering them. Is there any solution ?