In this chapter, we discussed some of the things we can do to improve the quality of our models, such as hyperparameter tuning to find the value or the combination of values among a set of candidate values that will give the best performance in our model. We also looked at how good defaults are important to starting experimentation with other candidate values. We discussed cross-validation when performing hyperparameter tuning and how it is important to leave a test set untouched so you can properly evaluate the results of the optimized model. Failing to do this can actually lead to adjusting the hyperparameters to the test set so that, depending on the applications, small improvements in performance could have great business impacts.
We also learned how some transformations to the target feature could improve the model. We used a logarithmic transformation to address...