Volvo_Discovery_Challenge_ECML_PKDD_2024

Implementation of the 2nd leaderboard submisison to the 2024 ECML PKDD Discovery Challenge

Predictive Maintenance for Volvo Trucks Using LSTM and Pseudo-Labeling

This repository contains the code for the methodology presented in the paper titled "Achieving Predictive Precision: Leveraging LSTM and Pseudo Labeling for Volvo’s Discovery Challenge at ECML-PKDD 2024". The paper details our approach to using Long Short-Term Memory (LSTM) networks and pseudo-labeling for predicting maintenance needs in Volvo trucks, which secured us the second place at the Volvo Discovery Challenge.

Overview

The methodology involves preprocessing the dataset to closely mirror the structure of the test set, using a base LSTM model to iteratively label the test data, and refining the model's predictions through a series of boosting techniques and pseudo-labeling. This approach enhanced the predictive capabilities of our models, achieving a macro-average F1-score of 0.879.

Repository Structure

This repository is organized into three main directories:

• EDA: Contains Jupyter notebooks and scripts used for exploratory data analysis (EDA). This includes inspection, visualization, and preliminary analysis of the dataset provided for the challenge.
• Model: Includes Python scripts for building and training the LSTM model. This section also contains the pseudo-labeling implementation and the boosting techniques used.
• Test and Evaluation: Features scripts used for testing, evaluation, and post-processing adjustments. This also includes consistency checks and scripts for generating final predictions and evaluating them against the provided metrics.

Citation

If you use the methodologies or the codebase in this repository in your research, please cite it as follows:

@inproceedings{volvo2024predictive, title={Achieving Predictive Precision: Leveraging LSTM and Pseudo Labeling for Volvo’s Discovery Challenge at ECML-PKDD 2024}, author={Carlo Metta et alt.}, booktitle={ECML-PKDD 2024}, year={2024} }

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

Contact

For any queries related to this project, please contact:

Carlo Metta, ISTI-CNR, Italy - carlo.metta@isti.cnr.it