Dataverse is an 8-hour datathon organised by BMSCE, Bangalore where participants analyze data from energy audits to identify inefficiencies and suggest improvements. Our team aimed to build a machine learning model that predicts the Energy Efficiency Rating of buildings and provide actionable recommendations for reducing energy wastage. This aligns with SDG-7: Clean and Affordable Energy.
- Predict Energy Efficiency: Use machine learning models to classify buildings into energy efficiency rating categories (A, B, C, D).
- Identify Inefficiencies: Highlight buildings with high energy consumption or inefficiencies.
- Generate Insights: Provide actionable recommendations for improving energy efficiency, such as reducing peak hour consumption or improving insulation.
- Aditya Ranjan
- Gnanendra Naidu N
-
Data Preprocessing: Cleaning, handling missing values, normalization, and feature engineering.
-
Machine Learning Models:
- Best Results:
- K-Nearest Neighbors
- Linear Discriminant Analysis
- Ridge Classifier
- XGBoost
- Suggestion Models:
- Qwen 32B
- GPT-4.0
- Best Results:
-
Evaluation Metric: F1-Score to balance precision and recall across energy efficiency ratings.
-
Best Results Models:
- K-Nearest Neighbors
- Linear Discriminant Analysis
- Ridge Classifier
- XGBoost
-
Position: Our team secured Third Place in the competition.
- Correlation Analysis:
Explored relationships between energy consumption, renewable usage, peak hours, floor area, and occupants. - Actionable Recommendations:
- Reduce peak hour consumption.
- Improve insulation for buildings with high energy inefficiency.
- Increase renewable energy utilization.
- BMS_Datathon_Dataverse.ipynb: Implementation of multiple ML models and suggestions using LLMs.
- Analysis_Correlation.ipynb: Detailed analysis of correlations between features.
- Dataverse.ipynb: Refinement of models by dropping less impactful parameters like floor area.
This project is licensed under the GNU General Public License v3.0.