Energy Consumption Forecasting & Analysis Project

In this project, I explored and predicted energy usage patterns using a dataset of nearly 20,000 hourly records. The data included temperature, humidity, occupancy, HVAC and lighting usage, and time-based variables like weekday and holiday status.

After cleaning and preparing the data (i.e. renaming columns, encoding categories, and correcting holiday labels), I performed detailed exploratory analysis. This revealed strong links between energy consumption and features like HVAC usage, occupancy, and temperature.

I trained several models, with the Random Forest Regressor performing best, achieving:

MAE: 0.23

RMSE: 0.38

R² Score: 0.90 on the test set

I used techniques like Robust Scaling, feature engineering, and cross-validation to improve performance and generalizability. Tools included R, dplyr, ggplot2, and caret.

This project helped uncover the key drivers of energy usage and built a solid foundation for smarter energy forecasting.