Explainable AI for Railway Track Maintenance: Integrating SMOTE and SHAP in Class‐Imbalanced Predictive Models.
Saved in:
| Title: | Explainable AI for Railway Track Maintenance: Integrating SMOTE and SHAP in Class‐Imbalanced Predictive Models. |
|---|---|
| Authors: | Sayed, Muhammad Masoom Mustafa1 (AUTHOR) sayed.2056619@studenti.uniroma1.it, Rind, Touqeer Ali2,3 (AUTHOR), Ahmed, Shiraz2 (AUTHOR), Khan, Muhammad Arsalan4 (AUTHOR), Mangnejo, Dildar Ali3 (AUTHOR), Shah, Muhammad Faarid2 (AUTHOR), Angiulli, Giovanni (AUTHOR) giovanni.angiulli@unirc.it |
| Source: | Journal of Engineering (2314-4912). 6/25/2026, Vol. 2026, p1-18. 18p. |
| Subjects: | Railroad track maintenance & repair, Shapley Additive Explanations, Artificial intelligence, Resampling (Statistics), Prediction models, Machine learning |
| Abstract: | This study presents an application‐oriented, interpretable evaluation of six supervised machine learning classifiers: gradient boosting, XGBoost, random forest, logistic regression, support vector machine, and artificial neural networks for railway track‐geometry defect prediction under class imbalance. The contribution is not the proposal of a new learning algorithm; rather, it is a domain‐specific assessment of how SMOTE‐based resampling changes minority‐defect detection and how SHAP explanations can make the resulting model behavior more auditable for maintenance planning. Using real inspection and defect records, the models are compared with emphasis on recall, F1‐score, and ROC‐AUC rather than accuracy alone. The results show that SMOTE improves minority‐class recall for several classifiers but also introduces precision trade‐offs, and overall discrimination remains moderate. SHAP analysis identifies defect length, directional traffic measures, speed variables, and tonnage‐related features as influential predictors. The findings therefore support the use of transparent predictive models as decision‐support tools for prioritizing railway maintenance, while recognizing the need for external validation before operational deployment. [ABSTRACT FROM AUTHOR] |
| Copyright of Journal of Engineering (2314-4912) is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | This study presents an application‐oriented, interpretable evaluation of six supervised machine learning classifiers: gradient boosting, XGBoost, random forest, logistic regression, support vector machine, and artificial neural networks for railway track‐geometry defect prediction under class imbalance. The contribution is not the proposal of a new learning algorithm; rather, it is a domain‐specific assessment of how SMOTE‐based resampling changes minority‐defect detection and how SHAP explanations can make the resulting model behavior more auditable for maintenance planning. Using real inspection and defect records, the models are compared with emphasis on recall, F1‐score, and ROC‐AUC rather than accuracy alone. The results show that SMOTE improves minority‐class recall for several classifiers but also introduces precision trade‐offs, and overall discrimination remains moderate. SHAP analysis identifies defect length, directional traffic measures, speed variables, and tonnage‐related features as influential predictors. The findings therefore support the use of transparent predictive models as decision‐support tools for prioritizing railway maintenance, while recognizing the need for external validation before operational deployment. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 23144904 |
| DOI: | 10.1155/je/1944007 |
