Disinformation-aware decision-making in supply chains using support vector machines
Abstract Supply chain decision-making is increasingly vulnerable to disinformation—false or misleading information that triggers demand forecasting errors, inventory mismanagement, and procurement disruptions. This study presents a systematic evaluation of five machine learning classifiers—support vector machine (SVM), logistic regression (LR), Naive Bayes (NB), random forest (RF), and XGBoost—for automated supply chain disinformation detection. A hand-verified dataset of 141 labeled statements spanning 12 supply chain topic areas was constructed from professional fact-checking organizations. All classifiers were evaluated using nested stratified 5-fold cross-validation with TF-IDF feature extraction. Linear classifiers substantially outperform ensemble tree methods: LR achieves the highest F1-score (82.2%) and precision (89.0%); SVM achieves the highest AUC-PR (0.925) and ROC-AUC (0.918); NB achieves the highest recall (84.2%). RF and XGBoost lag on F1 (72.4% and 72.5%). A Friedman test (χ 2 (4) = 7.435, p = 0.115) finds no statistically significant difference at the current dataset size; however, Cohen's d effect sizes up to 2.16 indicate a practically meaningful advantage for linear classifiers. A sensitivity analysis on the 104-row supply-chain-only subset confirms results are robust to source-domain confounding. SVM is recommended when threshold flexibility is required; LR when minimising false positives is the priority.
Authors
- Shweta Dwivedi (ORCID: https://orcid.org/0009-0000-7025-0564)
- Amita Nandal
- Saurav Kumar Jindal
- Bably Dolly
- Vijay Kumar
- Neeraj Kumar
Publication Details
- Journal
- Scientific Reports
- Published
- 2026-11-15
- DOI
- https://doi.org/10.1038/s41598-026-68744-8
- Primary Topic
- Misinformation and Its Impacts
- Type
- article
- Field-Weighted Citation Impact
- 0.00