Predicting the energy efficiency of buildings

In this section, we will investigate how neural networks can be used to solve a real-world regression problem. Once again, we turn to the UCI Machine Learning Repository for our data set. We've chosen to try out the energy efficiency data set available at http://archive.ics.uci.edu/ml/datasets/Energy+efficiency. The prediction task is to use various building characteristics, such as surface area and roof area, in order to predict the energy efficiency of a building, which is expressed in the form of two different metrics—heating load and cooling load.

This is a good example for us to try out as we can demonstrate how neural networks can be used to predict two different outputs with a single network. The full attribute description of the data set is given in the following table:

Column name	Type	Definition
relCompactness	Numerical	Relative compactness
surfArea	Numerical	Surface area
wallArea	Numerical	Wall area
roofArea	Numerical	...