ramonsanchez0213/urban-bus-ml-pipeline: Initial release

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Ramón Sánchez
Format: Recurso digital
Veröffentlicht: Zenodo 2025
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866902057124888576
author Ramón Sánchez
author_facet Ramón Sánchez
contents <p>This release contains the full implementation of the machine learning pipeline proposed in the study: "A Reproducible Pipeline for Leveraging Operational Data through Machine Learning in Digitally Emerging Urban Bus Fleets".</p> <p>The pipeline is designed for predictive maintenance in public transportation fleets that operate under data-scarce conditions, where limited historical records and fragmented infrastructure challenge the adoption of data-driven strategies.</p> <p>This repository includes:</p> <p>Domain-informed feature engineering tools</p> <p>Lightweight and interpretable ML models (Linear Regression, Ridge, KNN, Decision Tree)</p> <p>SMOGN for handling imbalanced regression targets</p> <p>Leave-One-Out Cross-Validation (LOOCV) for robust evaluation</p> <p>An adaptive batch retraining strategy for evolving data</p> <p>The pipeline has been validated using real operational data from hybrid diesel buses, focusing on predicting the time spent in critical soot accumulation zones of the Diesel Particulate Filter (DPF).</p> <p>Results show that while the model remained accurate in Zone 4 over time, retraining was necessary for Zone 3—highlighting the pipeline's ability to detect performance drift and adapt accordingly.</p> <p>This release is intended to support researchers, engineers, and practitioners working on predictive maintenance and digital transformation in public transport fleets.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_15874833
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle ramonsanchez0213/urban-bus-ml-pipeline: Initial release
Ramón Sánchez
<p>This release contains the full implementation of the machine learning pipeline proposed in the study: "A Reproducible Pipeline for Leveraging Operational Data through Machine Learning in Digitally Emerging Urban Bus Fleets".</p> <p>The pipeline is designed for predictive maintenance in public transportation fleets that operate under data-scarce conditions, where limited historical records and fragmented infrastructure challenge the adoption of data-driven strategies.</p> <p>This repository includes:</p> <p>Domain-informed feature engineering tools</p> <p>Lightweight and interpretable ML models (Linear Regression, Ridge, KNN, Decision Tree)</p> <p>SMOGN for handling imbalanced regression targets</p> <p>Leave-One-Out Cross-Validation (LOOCV) for robust evaluation</p> <p>An adaptive batch retraining strategy for evolving data</p> <p>The pipeline has been validated using real operational data from hybrid diesel buses, focusing on predicting the time spent in critical soot accumulation zones of the Diesel Particulate Filter (DPF).</p> <p>Results show that while the model remained accurate in Zone 4 over time, retraining was necessary for Zone 3—highlighting the pipeline's ability to detect performance drift and adapt accordingly.</p> <p>This release is intended to support researchers, engineers, and practitioners working on predictive maintenance and digital transformation in public transport fleets.</p>
title ramonsanchez0213/urban-bus-ml-pipeline: Initial release
url https://doi.org/10.5281/zenodo.15874833