Project-final

Reasearch Question

Which regression approach provides the most accurate forecasts of Switzerland’s national electricity consumption: Linear regression or Machine Learning models such as Lasso, Random Forest, and XGBoost?

Create Setup and envrironement

Steps to Run the Project

In a new directory of your choice, follow these steps:

1. Clone the project from GitHub

Open a terminal and run:

git clone https://github.com/mel799/Project-final.git
cd Project-final

2. Create a new Conda environment

conda env create -f environment.yml -n electricity-project

3. Activate the environement

conda activate electricity-project

Usage

4. Run the project

python main.py

All required datasets are already included in the repository in the following directory:data/raw/ No additional data download is required

Expected output:

• Loaded data
  • 2 folders are created in the data folder, "clean" and "dataset"
• EDA analysis
• Trained regression models
• Performance comparaison using regression metrics (RMSE, MAE, R²)
  • Evaluation results of the best regression model are stored in results/models/model_scores.txt
• Best Model visualization
  • stored in results/figures

Project Structure

Project-final/
├── main.py                 # Main entry point
├── src/
│   ├── data_loader         # Load data
│       └──GDP_monthly.py   # transform data to monthly
│       └──load_gdp.py       # loand and clean the GDP data
│       └── load_population.py
│       └── ...
│   ├── data_preprocessing  
│       └── add_features    # add features to the dataset
│       └── merge_data      # creates the dataset file
│       └── temp_heat_need  # adds the head feature
│   └── evaluation
│       └──eda_analysis         #analysis the dataset and patterns
│   └── model_visualisation.py  # creates graphics visualization
│   └── models_evaluation.py    # evalutates all the models with metrics
│   └── models_regression.py    # Regression models
├── test
│   └── random_foreset_tuning.py    #tune the random forest regression
├── data/
│   └── raw/
│       └── climate_Swiss.txt       # Swiss meteo data
│       └── electricity.csv         # Switzerland electricity consumption data
│       └── GDP_quarterly.cvs       # Swiss GDP quarterly
│       └── population_data_yearly  # population per year
└── environment.yml         # conda environment to download
└── PROPOSAL.md            # project proposal submitted in November
├── Project_report.tex      # Final report (LaTeX)
├── project_report.pdf     # Final report (pdf)
└── README.md               # Project documentation

Models Compared

• Linear Regression
• Lasso Regression
• Random Forest Regressor
• XGBoost Regressor

Results

• Best model: XGBoost RMSE: 107.10 MAE : 63.80 R2 : 0.9558

Requirements

• Python 3.11
• scikit-learn, pandas, matplotlib, seaborn, numpy, jupyter, xgboost, pyyaml