Any predictive model needs to be validated in different datasets to observe its behaviour. While the effectiveness of the model is usually kept constant over the data set with which we train it, it tends to vary depending on the sources from which the data come depending on how similar they are. To check if we have an overfitting problem, we must test if the model fits well to the data set we have provided to train it.
This overfitting problem is so common that there are multiple cross validation techniques, as we have seen in previous sections.
We are going to use this type of techniques to validate our predictive model on the “swiss” dataset. You can find an example of each of the techniques we have seen in the class notebook.
4. Stepwise Selection