Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{210631,
        author = {Zainab Imran Shaikh and Mehak Chainsingh Patel and Prof.Bharti Ganesh Salve},
        title = {CHAINGUARD AI: AN IoT-ENABLED FRAMEWORK FOR COLD-CHAIN SHIPMENT MONITORING, RISK ESTIMATION AND ANOMALY DETECTION},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {6},
        pages = {1691-1697},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=210631},
        abstract = {Cold-chain logistics requires continuous visibility of environmental conditions because temperature excursions, excessive humidity, transit delays and equipment faults can compromise temperature-sensitive goods. This paper presents CHAINGUARD AI, a prototype framework integrating shipment monitoring, machine-learning-based risk estimation, anomaly detection, rule-based alerts, route comparison and what-if scenario analysis. The supplied software uses a React and TypeScript frontend, a FastAPI backend, SQLAlchemy persistence, a Random Forest classifier and an Isolation Forest detector. The model feature schema includes temperature, relative humidity, transit duration, delivery delay, temperature exposure and recent temperature change. 
A predicted positive-class probability is mapped to a 0–100 prototype risk score and four display bands. Source-code inspection confirms that the current v0.1.0 model labels, seeded shipment records, telemetry and route inputs are synthetic; the supplied archive does not contain ESP32 firmware or a live DHT22 ingestion endpoint. To demonstrate the analytical workflow without presenting fabricated field evidence, a separate reproducible synthetic benchmark was generated using the same six-feature schema and a synthetic event-generation rule. 
On a stratified held-out synthetic test partition, the Random Forest achieved 85.60% accuracy, 53.44% recall, 56.45% F1-score and 0.811 ROC-AUC; logistic regression achieved 82.67% accuracy, 61.83% recall, 55.48% F1-score and 0.836 ROC-AUC. These values are simulation-only and cannot be interpreted as real shipment prediction performance. The paper documents the architecture, implementation, synthetic experiment and validation requirements for subsequent real-world testing.},
        keywords = {Cold-chain logistics, Internet of Things, shipment monitoring, Random Forest, Isolation Forest, anomaly detection, risk assessment, predictive analytics.},
        month = {November},
        }