Section 3 – Modeling Continuous Targets with Supervised Learning

The final ten chapters of this book introduce a wide range of machine learning algorithms, for predicting both continuous or categorical targets, or when there is no target. We explore models for continuous targets in this chapter.

A persistent theme in these chapters is that finding the best possible model is partly about balancing variance and bias. When our models fit the training data too well, they may not be as generalizable as we need them to be. In cases like that, they may have low bias but high variance. For each algorithm we examine in these chapters, we discuss strategies for achieving this balance. These strategies range from regularization for linear regression and support vector regression models, to the value of k for k-nearest neighbors, to the maximum depth of decision trees.

We also get a chance to practice the preprocessing, feature selection, and model evaluation strategies we worked...