The definition of overfitting is "the production of an analysis that corresponds too closely or exactly to a particular set of data, and may therefore fail to fit additional data or predict future observations reliably" (the model is good at training data and it is bad at test data).
But, is there a way to define overfitting programmatically ? For example; if a classification model's accuracy/f1 score is between %99 and %90 at training data and the model's accuracy/f1 score is equal or less than %80 at test data, the model overfits. Or if a regression model's rmse value is equal or less than 0.7 at training data (target variable ranges from 0 to 1000) and the model's rmse value is equal or more than 5.0 at test data, the model overfits.
