Research work aims at introducing a new estimation method for statistical machine learning model-fit to Test-Data without Test-Data by applying Generalized AIC (Akaike Information Criterion)-Takeuchi 1976 and Empirical risk information criteria. Research work is carried under the supervision of Dr. Richard Golden. We have compiled noteworthy results for basic machine learning model, currently we are generalizing the estimation process to more multifaceted neural network model based on deep learning and using non-parametric estimation approaches on various dataset (Bootstrapping with replacement), to ascertain the generalization of the proposed method.