This repository was developed as part of the Brick by Brick 2024 Challenge, a global competition aimed at automating building data classification to enhance the intelligence and energy efficiency of buildings. By tackling this challenge, we contribute to the broader effort of enabling sustainable and smart building management through data-driven solutions.

Our machine learning pipeline achieves a 0.5767 macro F1 score on the public test leaderboard. The approach leverages advanced feature extraction and ensemble learning to optimize performance. For a detailed explanation of the methodology, feature extraction process, and model optimization, please refer to the accompanying report.pdf and our published paper.

The Brick by Brick 2024 Challenge is a global initiative designed to advance automation in building data classification. The challenge's primary goal is to enable intelligent, energy-efficient buildings by leveraging machine learning and data science methodologies. Our contribution focuses on developing a robust, high-performing model that effectively processes and classifies building data.

The solution is built around feature extraction and ensemble learning, leveraging a structured approach to feature selection and hyperparameter optimization. The pipeline follows a stratified k-fold cross-validation scheme to ensure robustness.

• model.py: Contains the code to train the model based on extracted features. It also includes the feature selection paradigm and hyperparameter optimization using a stratified k-fold approach.
• brick_feature_extractor.py: Implements feature extraction, generating time-based, interval-based, and full dataset features, which are stored in the features/ folder for reference.
• dataloader.py: Merges extracted features for both training and test datasets.
• Utility Files: Contains schema definitions and available classes relevant to the brick classification problem.
• submission/: Includes all solution submission files related to the paper.

If you use this code or refer to our work, please cite the following paper:

@inproceedings{Steenwinckel2025,
  author    = {Bram Steenwinckel and Sofie Van Hoecke and Femke Ongenae},
  title     = {Another Brick in the Wall: Leveraging Feature Extraction and Ensemble Learning for Building Data Classification},
  booktitle = {Companion Proceedings of the ACM Web Conference 2025 (WWW Companion '25)},
  year      = {2025},
  publisher = {ACM},
  doi       = {10.1145/3701716.3718480},
  isbn      = {979-8-4007-1331-6/2025/04},
}

This code was executed using Python 3.11 with the following necessary packages:

• pandas (latest version)
• scikit-learn (latest version)

This code is licensed under the Apache License 2.0. See the LICENSE file for details.