Fine Tuning (Diagnosis) of Machine Learning Algorithm (Model) for optimization

When any machine learning algorithm is deployed/tested on new set of unseen data, and makes unacceptable large error in its prediction then it became mandatory to diagnose present hypothesis and evaluate it by implementing the one, some or all of the following a) get more training data. b) implementing smaller set a features. c) implementing additional features. d) adding polynomial features. e) decreasing or increasing regularization parameter λ. Machine learning diagnosis is a way to find out what is working and what not for an algorithm. The developer must determine what may be changed to increase the performance of an algorithm on unseen set of data generated in future. Diagnosis of any machine can take time to understand and implement but, they can also enable machine for the better decision making.