Want to add predictor output results of a model to the training set of the same model.

I have large set of molecules. I have partitioned them in 30:70 and build a model initially and predict the 70% of the molecules. Then I want to add the few selected molecules after their prediction into the training set of same model. That process will be iteratively run 10-15 times. But the problem is whenever I add the output of the predictor set to the training set, the model will be changed so the predictor result is going to vanished and it is quite obvious though. So can anyone help me in this issue?

Thank you in advance.

Sourav.