In Part 1 of this two-part series, I examined some of the ways the artificial intelligence (AI) algorithms that are used in HVAC control systems use data to approximate functions. In a simplistic approach, the AI algorithms used in our industry could be grouped into two types: classification and regression, both of which fall under the category of supervised learning, i.e., learning to approximate a function by “seeing” the data. Classification algorithms are used to predict discrete values, i.e., true or false.
This article will focus on regression-type algorithms. Some examples of regression-type algorithms include simple linear regression, multiple linear regression, polynomial regression, support vector regression, decision tree regression, and random forest regression. These algorithms are used to predict continuous variables, i.e., chiller plant energy consumption, cooling coil leaving air temperature, etc. Each one of these variables can be approximated as a function of a single input variable or multiple input variables.