Regularization: Machine Learning

For understanding the concept of regularization and its link with Machine Learning, we first need to understand why do we need regularization. We all know Machine learning is about training a model with relevant data and using the model to predict unknown data. By the word unknown, it means the data which the model has not seen yet. We have trained the model, and are getting good scores while using training data. But during the process of prediction, we found that the model is underperforming when compared to the training part. Now, this may be a case of over-fitting(about which I will be explaining below) which is causing incorrect prediction by the model.