Overview

The Energy Efficiency Prediction project integrates MLflow for experiment tracking and applies machine learning techniques to forecast heating and cooling loads in buildings. It emphasizes the utilization of architectural features to enhance the understanding of energy efficiency and to increase the accuracy of model predictions.

Repository Contents

• ENB2012_data.csv: Dataset with architectural features and target variables.
• notebook.ipynb: Jupyter notebook with the project's code and analysis.
• mlruns/: MLflow logs and registered models.

Methodology

• Exploratory Data Analysis: Involves the loading of data and conducting EDA using histograms, scatter plots, and heatmaps.
• Feature Optimization: Focuses on enhancing the dataset by implementing strategic feature engineering, aiming to refine and utilize the most relevant information.
• Data Preprocessing: Involves preprocessing of data, which includes tasks like standardizing data and splitting it into training and testing sets.
• Model Training and Evaluation: Involves the development and assessment of various regression models, such as Decision Trees, Random Forests, and Gradient Boosting
• Key Feature Identification: Conducts a thorough analysis to pinpoint and emphasize critical features impacting model predictions, like 'Relative Compactness' and 'Overall Height'.
• Performance Assessment: Focuses on evaluating the models based on key performance indicators, including RMSE, MAE, and R².
• Utilization of MLflow: Employing MLflow for comprehensive experiment tracking, efficient model management, and ensuring streamlined deployment of models to production environments.

Installation and Usage

To set up the project, clone the repository, install dependencies, and navigate to the directory. Run Jupyter Notebook to view the project:

git clone https://github.com/Shanmukhi1920/Energy_Efficiency
cd Energy_Efficiency
pip install -r requirements.txt
jupyter notebook

Then, open Notebook.ipynb in Jupyter.

• Access trained models and logs in the mlruns/ directory.

Key Findings

Model Performance Overview: The Gradient Boosting model exhibits strong performance and generalization for both heating (HL) and cooling load (CL) predictions, with R-squared values near 1.0 across all datasets, indicating nearly perfect fit and prediction accuracy.

Heating Load Predictions: Heating load predictions are highly accurate with a training RMSE of 0.27, validation RMSE of 0.45, and test RMSE of 0.427, demonstrating the model's precise fit and consistent performance across different data sets.

Cooling Load Predictions: Cooling load predictions, while accurate, show slightly higher errors with a training RMSE of 0.39, validation RMSE of 0.97, and test RMSE of 0.786, indicating a need for further refinement to achieve the precision seen in heating load predictions.

Consistency and Robustness: The model demonstrates robustness and consistency, with test R-squared values of 0.998 for HL and 0.993 for CL, suggesting strong predictive capabilities and reliability for practical applications.