Document Type
Student Coursework
Publication Date
Fall 11-1-2026
Abstract
Global semiconductor manufacturing networks are characterized by deep geographical concentration, high capital intensity, and acute vulnerability to regional supply disruptions. While public merchandise trade statistics provide transparent, reproducible empirical records, transforming these raw macroeconomic streams into actionable supply chain risk intelligence requires both rigorous analytical framing and scalable computational architecture. In this paper, we propose an enterprise-grade cloud data architecture designed for semiconductor risk monitoring, combining scheduled distributed batch processing on Google Cloud (Dataproc/PySpark, BigQuery) with an event-driven streaming ingestion lane (Pub/Sub, Dataflow/Apache Beam). To rigorously evaluate the underlying scoring and forecasting logic without premature distributed deployment costs, we implement and validate a complete local pipeline processing 20,303 monthly U.S. Census trade records spanning January 2023 through March 2025 across Harmonized System (HS) commodity codes 8486, 8541, and 8542. The analytical engine establishes parametric, multi-attribute exposure formulations for sovereign supplier dependency and product-level concentration, corroborated through multi-scenario weight sensitivity tests. Across all four scenario variations, Malaysia consistently leads sovereign risk (baseline score 82.7), driven by substantial trade share and elevated short-term volatility, while HS 8542 (electronic integrated circuits) dominates product risk (50.4). Furthermore, recursive short-term forecast models selected via rolling out-of-sample backtests achieve an aggregate Mean Absolute Percentage Error (MAPE) of 8.57% across nine strictly held-out product-month actuals (April–June 2025). We demonstrate that while point forecasts remain numerically accurate, discrete disruption alert classifications are highly sensitive to threshold proximity, arguing for multi-metric evaluation over single-point MAPE. Finally, we provide a structured architectural assessment contrasting custom distributed data pipelines against low-code/AI workflow platforms.
Program or Discipline Name
Computer and Information Sciences
Recommended Citation
Gong, M., & Cha, S. (2026). Semiconductor Supply Chain Risk Monitoring and Forecasting: A Hybrid Big Data Architecture and Empirical Validation. Semiconductor Supply Chain Risk Monitoring and Forecasting: A Hybrid Big Data Architecture and Empirical Validation, 1-13. Retrieved from https://digitalcommons.harrisburgu.edu/other-works/35
Publication Title
Semiconductor Supply Chain Risk Monitoring and Forecasting: A Hybrid Big Data Architecture and Empirical Validation
Start Page No.
1
End Page No.
13
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Submission aggreement