Notes

Hierarchical clustering to separate the dataset into groups


EDA and data cleaning:
- Convert ints to categorical variable where required
- Use one-hot encoding to divide up the data set based on glazing area, and glazing area distribution
- 2 factors with 4 states each
  - F-test against the zero glazing, zero orientation base case
  - If there is no significant difference, drop certain cases 

Classification
- Predict energy efficiency category (high, medium, low) based on glazing area, height, and distribution


Scale the variables to give balance to the results
Classification for making predictions on the training set
Regression to predict values