Predictive Modeling of Building Energy Loads
Linear regression and k-means clustering to predict heating and cooling loads in residential buildings.
Repository: pratham-aggr/energy_loads
This project applied supervised and unsupervised machine learning to predict the heating and cooling loads of residential buildings from architectural features (wall area, roof area, glazing area, orientation, etc.).
Methods:
- Linear regression — achieved 91% accuracy (R²) on held-out test data for both heating and cooling load prediction.
- k-means clustering — clustered buildings by energy profile to identify groups with similar load characteristics, providing interpretable structure beyond point predictions.
The dataset originates from UCI Machine Learning Repository energy efficiency data. The work demonstrates the effectiveness of simple, interpretable models for energy efficiency prediction — relevant to sustainable building design and climate-aware construction.