Minimizing the predicted output from a DNN multiple regression model

Question

I have trained a DNN to perform multiple linear regression (MLP) based on data from 9 input variables. The model works well and predicts the single output variable in a sati