Unicatt: Power Load Forecasting

This repository contains code and resources for power load forecasting using machine learning techniques. The project leverages historical load data, and calendar information to predict future power consumption patterns.

Setup Local Environment

1. Clone the repository:
   git clone https://github.com/xtreamsrl/unicatt-2025-forecasting.git
2. Install uv:
   curl -LsSf https://astral.sh/uv/install.sh | sh
3. Check uv installation:
   uv --version
4. Install uv dependencies:
   uv sync

Project Structure

• data/: Contains datasets used for training and testing the models.
• notebooks/: Jupyter notebooks for exploratory data analysis and model development.
• src/: Source code for data preprocessing, model training, and evaluation.
  • src/models/: Saved machine learning models and relative configurations.
  • src/preprocessing/: Scripts for feature engineering.

Learning Objectives

This project demonstrates best practices for building production-ready machine learning systems through:

1. Problem-Solving Methodology: Learn how to approach forecasting problems systematically by:

   • Conducting exploratory data analysis on time series data
   • Iterating through models of increasing complexity

2. Code Architecture for Production: Understand how to structure ML codebases for maintainability and extensibility by:

   • Organizing code into reusable components (data preprocessing, model training, evaluation)
   • Writing modular code so you can swap out preprocessing steps or models without rewriting the entire pipeline