Devina707/Building-Energy-Demand-Prediction

⚡ BuildSmart — Energy Demand Predictor

ML-powered building energy meter reading prediction using HistGradient Boosting, trained on the ASHRAE Great Energy Predictor III dataset (Site 1).

Features

• —📊 EDA tab — raw data, descriptive stats, building use, area, age, meter type, and correlation analysis
• —🏆 Model Performance tab — 5-model comparison, hypertuning results, feature importance, improvement suggestions
• —⚡ Predict tab — input building + weather details, get instant kWh prediction

Model

• —Algorithm: HistGradient Boosting Regressor (hypertuned)
• —Best params: learning_rate=0.1, max_depth=7, max_iter=200, min_samples_leaf=20, l2_regularization=0.1
• —R² = 0.9178 · MAE = 27.88 kWh · RMSE = 50.82 kWh

Repo Structure

app.py                          ← single entry point
requirements.txt
README.md
src/
  pipelines_inference_st.pkl    ← trained model pipeline
  df_raw.csv                    ← raw ASHRAE Site 1 data
  parameter-speedometer-energy-free-png.png

Built by Devina Agustina as part of the Hacktiv8 Milestone Project.