Chapter 7: Online Regression
After looking at online anomaly detection and online classification throughout the previous chapters, there is one large category of online machine learning that remains to be seen. Regression is the family of supervised machine learning models that applies to use cases in which the target variable is numerical.
In anomaly detection and classification, you have seen how to build models to predict categorical targets (yes/no and iris species), but you have not yet seen how to work with a target that is numerical. Working with numerical data requires having methods that work differently, both in the deeper layers of model training and model definition and also in our use of metrics.
Imagine being a weather forecaster trying to forecast the temperature (Celsius) for tomorrow. Maybe you expect a sunny day, and you have a model that you use to predict a temperature of 25 degrees Celsius. Imagine if the next day, you observe that it is cold and only 18...
