We used a wireless sensor network, which measured:

• the temperature in different rooms of the house
• humidity in different rooms of the house
• temperature outside
• humidity outside
• wind outside 

We know the electric consumption.

We transformed the data to fit a classification problem.

We grouped the samples by intervals of energy consumption. Each one being a class in the final dataset.

So for example, between 0 and 200Wh belongs to class 0, between 200 and 400 to class 1 etc.

We have a total of 5 classes.

After training, we are able to predict future energy use.