INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 403
www.rsisinternational.org
AI and Predictive Analytics for Supply Chain Risk Management:
Opportunities for U.S. Manufacturing Resilience
Nasima Akter
Atlantis University, Miami Florida
DOI:
https://doi.org/10.51583/IJLTEMAS.2026.15020000036
Received: 16 February 2025; Accepted: 21 February 2026; Published: 05 March 2026
ABSTRACT
Recent shocks in the global supply chain such as pandemic related shutdowns and semiconductor shortages have
shown the susceptibility of the U.S. manufacturing supply chains. Conventional risk management practices that
are mostly based on historical averages and reactive practices have not been effective in unpredictable
environments. This paper examines the connection between artificial intelligence (AI) and predictive analytics
and reinforcing risk identification, forecasting, and resilience in the U.S. manufacturing industry. We assess the
performance of predictive models including ARIMA, Prophet, and Random Forest using the secondary data
from U.S. Census Bureau, Bureau of Labor Statistics, and publicly available logistics datasets. The most
important measures of resilience are forecast error (MAPE, RMSE), inventory turnover, order fulfillment, and
recovery time. The findings indicate that predictive analytics can greatly reduce errors in predictions as well as
enhance the outcomes of resilience with the Random Forest being better in terms of reducing forecast error by
up to 30% as well as increasing order fulfillment rates by 15% compared to traditional models. A model-derived
association analysis using simulation-based performance metrics further indicates statistically significant
directional relationships between predictive analytics capability and resilience indicators. These results highlight
the value of predictive analytics as an input to operational decision-making. To managers, the study illustrates
the use of digital tools to help them get actionable information on the risks of disruption and to policymakers, it
is part of wider efforts to strengthen national supply chain resilience. This study can enhance the empirical
understanding of AI-facilitated models and introduce feasible resilience metrics.
Index terms: Artificial intelligence, Predictive analytics, Supply chain resilience, Risk management, U.S.
manufacturing, Forecast accuracy, Business intelligence.
INTRODUCTION
Supply chains supporting U.S. manufacturing have become increasingly vulnerable to large-scale disruptions
caused by pandemics, geopolitical tensions, and critical material shortages. The COVID-19 crisis exposed how
quickly global production and distribution networks can break down, particularly in strategically important
sectors such as semiconductors, automotive manufacturing, and consumer goods. These disruptions revealed the
limitations of traditional, reactive risk management approaches that rely heavily on historical averages rather
than forward-looking decision support. As a result, the ability to anticipate disruptions and respond proactively
has emerged as both an operational necessity for firms and a strategic priority for U.S. industrial policy.
Recent literature identifies supply chain resilience as a critical capability that enables firms to anticipate, absorb,
recover from, and adapt to unexpected shocks. While prior studies emphasize concepts such as agility,
redundancy, and flexibility, resilience measurement remains inconsistent across empirical research. Many
studies rely on qualitative frameworks or generalized indicators, limiting their ability to translate theoretical
insights into actionable, data-driven decision-making.
At the same time, advances in artificial intelligence (AI) and predictive analytics have created new opportunities
for proactive supply chain risk management. Time-series forecasting and machine-learning models have
demonstrated strong technical performance in demand prediction and anomaly detection. However, much of this
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 404
www.rsisinternational.org
research remains either technically focused or globally aggregated, with limited empirical examination of how
predictive analytics directly contribute to resilience outcomes within the U.S. manufacturing context.
This paper contributes by (1) providing U.S.-context evidence using a model-driven, secondary-data approach,
(2) operationalizing resilience through measurable indicators linked to predictive performance, and (3)
presenting a Predictive Resilience Framework that connects analytics capability to resilience outcomes in a
unified structure.
LITERATURE REVIEW
A. Supply Chain Risk Management and Resilience.
Supply chain risk management (SCRM) is a topic that has occupied the center of the attention of operations
research over a period of 20 years. Risks found in the early frameworks were attributed to the variation of
demands, supply interference, or network complexity (Juttner et al., 2003; Tang, 2006). The capacity of a system
to envision, absorb, recuperate and adjust to shocks, which is known as resilience has been identified as an
essential capacity of sustaining continuity in turbulent surroundings (Ponomarov and Holcomb, 2009). Recent
research emphasizes that resilience is not merely a matter of survival following the incident but a process that
prepares firms to gain a competitive edge by responding quicker than competitors (Ivanov, 2021).
B. Conceptual vs. Empirical Approaches:
Even though resilience is a well-discussed concept, a substantial part of the literature is still abstract. The
constructs of resilience that researchers tend to define include agility, flexibility, redundancy, but do not
operationalize them. A substantial body of research is based on surveys or qualitative assessment based on cases
(Ali et al., 2017), which offer useful information but are not suitable in terms of comparing results across
different settings. Empirical research based on quantifiable measures, e.g., forecast error, inventory turnover,
stockout risk or recovery time, is relatively limited with a gap in rigorous, evidence-based evaluation.
C. Analytics and Artificial Intelligence:
Simultaneously, the development of analytics has revolutionized the supply chain management. Descriptive
dashboards, business intelligence (BI) as well as early decision-support systems assisted managers to track
performance and detect risks. And more recently, predictive and prescriptive analytics, enabled by artificial
intelligence (AI) and machine learning (ML) provide means of proactive risk mitigation. ARIMA, neural
networks, ensemble learning, and Prophet are some of the techniques tried in demand forecasting and logistics
planning (Box et al., 2015; Breiman, 2001; Choi et al., 2018). These studies indicate technical potential of
predictive analytics, but most of them discuss it in terms of accuracy of forecasting, but not the resilience
outcomes.
D. Integration Gap: Analytics and Resilience:
A recurring limitation in existing research is the disconnection between technical model validation and
managerial outcomes. Many AI-focused studies evaluate models based on error reduction metrics (MAPE,
RMSE), but stop short of linking these improvements to resilience indicators such as reduced recovery time or
improved order fulfillment. Conversely, resilience studies often discuss strategies conceptually without adopting
modern predictive tools. This divide prevents the literature from offering integrated frameworks that both
validate model performance and show tangible resilience benefits.
E. Geographical and Contextual Limitations:
Another gap concerns the scope of empirical work. Much of the research on predictive analytics and resilience
is conducted in European or Asian contexts, where data availability and industrial cooperation are more
established. U.S. manufacturing, critical for economic competitiveness and national security, remains
underrepresented in the academic discourse. While policy reports emphasize the urgency of strengthening
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 405
www.rsisinternational.org
domestic supply chains, scholarly research has not yet provided comprehensive empirical evidence on how
predictive analytics can enhance resilience specifically in U.S. industries.
F. Research Gap and Contribution
In summary, the literature reveals three main gaps: 1. Limited measurement, resilience often lacks quantifiable
indicators, restricting empirical comparison. 2. Weak integration as predictive analytics studies rarely connects
model performance to resilience outcomes and 3. Contextual underrepresentation like U.S. manufacturing supply
chains have not been sufficiently examined.
This study addresses these gaps by: Developing a resilience measurement framework that incorporates predictive
model outputs into operational metrics. Comparing traditional forecasting models (ARIMA, Prophet) with
machine learning (Random Forest) to link accuracy improvements with resilience outcomes. Situating the
analysis within U.S. manufacturing, using both secondary data and simulation-based resilience evaluation.
By doing so, the research not only strengthens the empirical foundation of resilience studies but also provides
actionable insights for managers and policymakers navigating the uncertainties of modern supply chains.
RESEARCH METHODOLOGY
This study adopts a quantitative, model-driven research design to examine how predictive analytics enhance
supply chain resilience in U.S. manufacturing. The analysis is based on secondary industry data, predictive
modeling, and simulation rather than firm-level survey inference. Multiple forecasting models are evaluated and
linked to operational resilience indicators to assess the practical and theoretical implications of artificial
intelligence-enabled decision support.
A. Data Sources and Study Scope:
The analysis relies exclusively on publicly available secondary data related to U.S. manufacturing, automotive
production, semiconductor supply chains, and macroeconomic indicators covering the period 2015–2022. These
datasets capture demand variability, production output, inventory behavior, and disruption patterns relevant to
resilience analysis. The use of secondary data ensures transparency, reproducibility, and consistency across
sectors.
B. Predictive Modeling Approach
Three predictive models were implemented and compared: Autoregressive Integrated Moving Average
(ARIMA), Prophet, and Random Forest. ARIMA was selected as a baseline statistical forecasting method,
Prophet for its ability to capture seasonality and trend shifts, and Random Forest to represent machine-learning-
based nonlinear prediction. Models were trained using historical demand and production data and evaluated
using rolling forecasts to reflect operational decision-making conditions.
C. Resilience Indicators:
Supply chain resilience was operationalized using multiple performance indicators, including forecast accuracy
(measured by MAPE and RMSE), inventory turnover, stockout probability, order fulfillment rate, and recovery
time. These indicators capture both predictive performance and downstream operational outcomes, allowing
forecast quality to be directly linked to resilience behavior.
D. Simulation-Based Impact Assessment
To translate predictive performance into operational resilience outcomes, a simulation-based assessment was
conducted. Forecast outputs were used as inputs into simulated inventory and fulfillment scenarios under
disruption conditions. Percentage improvements reported in this study represent relative changes between
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 406
www.rsisinternational.org
baseline forecasting scenarios and AI-enhanced predictive scenarios, enabling evaluation of early-warning
capability, service continuity, and recovery speed.
E. Model-Derived Association Analysis
To examine the relationship between predictive analytics capability and resilience outcomes, a model-derived
association analysis was conducted. Regression-style coefficients were estimated using simulated performance
metrics rather than firm-level observations. The results illustrate the directional influence of predictive capability
on resilience indicators and support the conceptual relationships proposed in the Predictive Resilience
Framework.
F. Robustness Checks
Robustness checks were performed across multiple manufacturing sectors, including automotive,
semiconductor, and textile industries. Model performance and resilience outcomes were compared across sectors
to ensure that results were not driven by sector-specific dynamics. The relative superiority of machine-learning-
based predictions remained consistent across all tested contexts.
RESULTS
This section presents the empirical outcomes of the predictive modeling and simulation-based analysis. Results
are organized to first compare the forecasting performance of ARIMA, Prophet, and Random Forest models,
followed by an assessment of their implications for supply chain resilience indicators.
A model-derived association analysis is then reported to illustrate the directional relationships between
predictive analytics capability and resilience outcomes. Finally, a sector-specific case study is presented to
demonstrate the practical application of predictive analytics under disruption conditions.
A. Model-Derived Association Analysis
To examine the relationship between predictive analytics capability and supply chain resilience outcomes, a
model-derived association analysis was conducted using simulated performance metrics. The results illustrate
the directional influence of enhanced predictive capability on key resilience indicators rather than firm-level
causal inference.
TABLE I. Model-Derived Association Between Predictive Analytics Capability and Supply Chain Resilience
Indicators
Resilience Indicator
Coefficient (β)
Std. Error
t-value
Significance (p)
Forecast Accuracy (MAPE)
-0.312
0.087
-3.58
<0.01
Inventory Turnover
0.271
0.102
2.66
<0.05
Stockout Probability
-0.298
0.091
-3.28
<0.01
Recovery Time (days)
-0.245
0.110
-2.23
<0.05
Order Fulfillment Rate
0.321
0.095
3.37
<0.01
Note: Coefficients are derived from simulation-based performance metrics and illustrate directional relationships
rather than firm-level causal inference.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 407
www.rsisinternational.org
B. Predictive Model Performance
Three predictive models were tested: ARIMA, Prophet, and Random Forest. Performance was evaluated using
AUC, precision, recall, and forecasting error metrics. Results indicate that Random Forest outperformed
traditional models, reducing forecast error by approximately 30% compared to ARIMA and improving order
fulfillment by 15%.
Percentage improvements were computed relative to the ARIMA baseline. Forecast error improvement reflects
the percentage reduction in MAPE and RMSE, while order fulfillment improvement reflects the percentage
change in simulated fulfillment rates under AI-informed scenarios compared with the baseline scenario.
TABLE II. Predictive Model Performance Metrics
AUC
Precision
Recall
MAPE
RMSE
0.78
0.70
0.65
18.4%
4.12
0.85
0.74
0.71
15.9%
3.64
0.91
0.82
0.80
12.8%
2.87
Figure 1 compares the relative performance of the three predictive models in identifying disruption-related risk
signals, highlighting differences in early-warning capability.
Figure 1. Predictive Model Comparison Based on Classification Performance.
This figure compares the relative ability of ARIMA, Prophet, and Random Forest models to identify disruption-
related risk conditions using model-based performance metrics. Higher values indicate stronger early-warning
capability.
Figure 2 presents a comparative evaluation of forecast accuracy across models, measured using MAPE and
RMSE to capture both absolute and relative error behavior.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 408
www.rsisinternational.org
Figure 2. Forecast Error Comparison Across Predictive Models.
Forecast accuracy is evaluated using MAPE and RMSE across ARIMA, Prophet, and Random Forest models.
Lower error values indicate improved predictive performance under volatile demand conditions.
C. Robustness Checks
Robustness checks were conducted across multiple manufacturing sectors, including automotive, semiconductor,
and textile industries. While absolute performance levels varied by sector, the relative ranking of predictive
models remained consistent. Machine-learning-based predictions continued to outperform traditional statistical
models across forecast accuracy and resilience-related indicators, suggesting that the observed relationships are
not driven by sector-specific dynamics.
CASE STUDY
Auto Supply Chain in the U.S. in the Semiconductor Shortage
A. Context
The case study evaluates the operational implications of predictive analytics in the context of the U.S. automotive
supply chain during the semiconductor shortage. Model outputs were translated into simulated operational
scenarios to assess early-warning capability, inventory response, and service performance under disruption
conditions.
B. Predictive Models Usage.
Past demand and supply data in 2015-2022 were obtained using publicly available datasets of the automotive
industry. Three forecasting models, which included ARIMA, Prophet and Random Forest, were trained to make
the demand of the chips based on the expected automotive output. One of the criteria used to evaluate the models
was the capability to predict shortages and make inventory decisions.
ARIMA was effective in the long-term demand trend, but it was not effective in sharp and abrupt volatility,
which gave high forecast errors when disruptions are at their peak levels.
The use of Prophet also did better at capturing seasonality and other external regressors and yet overestimated
the magnitude of shortages.
Random Forest can also identify initial signs of shortage risk more accurately since it used several predictors
such as international trade flows and macroeconomic variables.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 409
www.rsisinternational.org
C. Case Study Simulation Assumptions:
Forecasts were generated on a rolling basis using historical demand proxies, production output indicators, and
macroeconomic variables. Shortage risk was flagged when forecasted chip demand exceeded a simulated supply
capacity threshold for a sustained period. Early warning lead time was defined as the time between the first
model signal crossing the risk threshold and the onset of the disruption period. Order fulfillment outcomes were
simulated using baseline versus AI-informed inventory adjustment scenarios to estimate relative changes in
fulfillment performance and recovery time.
D. Knowledge Advantages over the Conventional Methods:
The customary risk management within the sector has used supplier guarantees and reacted changes, which are
not sufficient. Conversely, the case study indicates that Random Forest minimized error in forecasting by 28
percent against ARIMA and forecasted possible shortages 3 months before. The lead time would have allowed
companies to develop buffer stock or diversify the sourcing strategies, prior to the shutdowns of production. The
order fulfillment rates, which were simulated using the predictive model, increased by 12 percent compared to
the base case.
The case study highlights the importance of predictive analytics in the context of the gap between monitoring
and proactive resilience. Although ARIMA and Prophet offer improved increments, machine learning systems
such as Random Forest are able to offer actionable predictability in excessively volatile supply settings. Notably,
this fact confirms the wider findings of the research process and proves the practical application of the
introduction of AI tools into the supply chains of the U.S. manufacturing industry.
DISCUSSION
A. Interpretation of Predictive Model Performance
The results demonstrate that machine-learning-based predictive analytics offer measurable advantages over
traditional statistical forecasting methods under volatile supply chain conditions. While ARIMA and Prophet
capture long-term trends and seasonality, their performance deteriorates when demand volatility increases. In
contrast, Random Forest models exhibit greater flexibility in incorporating multiple predictors and nonlinear
relationships, which explains their superior performance in forecasting accuracy and early risk detection. These
findings reinforce the value of advanced analytics for proactive supply chain risk management rather than
reactive response mechanisms.
B. Predictive Analytics as a Driver of Supply Chain Resilience:
The model-derived association analysis indicates that improved predictive capability is directionally linked to
multiple resilience indicators, including lower forecast error, reduced stockout probability, improved order
fulfillment, and shorter recovery time. These results support the conceptual argument that forecasting accuracy
is not an isolated technical metric but a foundational capability that enables operational resilience. By translating
predictive signals into simulated inventory and fulfillment decisions, the study demonstrates how analytics-
driven foresight can support continuity and faster recovery during disruption events.
This interpretation aligns with resilience literature that emphasizes anticipation and adaptability as core
resilience dimensions (Ponomarov and Holcomb, 2009; Ivanov, 2021).
C. Insights from the Semiconductor Shortage Case Study:
The case study of the U.S. automotive supply chain during the semiconductor shortage illustrates how predictive
analytics can be operationalized under real disruption conditions. Simulation results suggest that earlier detection
of shortage risk enables firms to adjust sourcing, inventory buffers, and production planning in advance of
disruption onset. This finding extends prior research on supply chain viability by demonstrating how predictive
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 410
www.rsisinternational.org
analytics can convert early-warning signals into actionable resilience responses rather than post-disruption
recovery actions (Chopra, 2020).
D. Theoretical Implications: Predictive Resilience Framework:
This study contributes to resilience theory by operationalizing the linkage between analytics capability and
resilience outcomes through the proposed Predictive Resilience Framework. Unlike conceptual resilience
models that emphasize flexibility or redundancy without measurable benchmarks, this framework connects
business intelligence adoption and AI-enabled prediction directly to quantifiable operational outcomes. By
grounding resilience in observable performance indicators, the framework advances resilience research toward
more empirical and decision-oriented models.
E. Managerial and Policy Implications:
From a managerial perspective, the findings suggest that investments in predictive analytics can enhance
resilience by enabling earlier risk detection and more informed operational decisions. For policymakers, the
results provide empirical support for initiatives that promote digital infrastructure, analytics capability, and
workforce upskilling in strategically important manufacturing sectors. These insights are consistent with broader
efforts to strengthen domestic manufacturing resilience through technology-enabled decision support.
F. Limitations and Future Research
This study relies on secondary data and simulation-based analysis, which enhances transparency and
reproducibility but limits access to firm-level operational constraints and proprietary decision processes. Future
research could extend this framework using firm-level datasets or hybrid modeling approaches that integrate
statistical forecasting with advanced machine-learning techniques across additional sectors critical to U.S.
manufacturing competitiveness.
CONCLUSION
This paper reveals that artificial intelligence and predictive analytics can make supply chains more resilient in
the U.S. manufacturing industry. Using predictive analytics to lower forecasting error, enhance order fulfillment,
and quicken recovery time, the work demonstrated that predictive analytics using ARIMA, Prophet, and Random
Forest models on sectoral data are more effective than conventional methods. The semiconductor shortage case
study also demonstrates the usefulness of AI-based tools in helping to detect the risks sooner and provide
proactive action.
There are three contributions of this paper. First, it offers empirical data in the context of the U.S., which fills
the gap in the resilience literature that has frequently focused on European and Asian supply chains. Second, it
enhances the operationalization of resilience through four indicators, accuracy of the forecast, stockout
likelihood, recovery time and fulfilment of orders, which relate the performance of the predictive model directly
to resilience results. Third, it suggests a Predictive Resilience Framework that combines business intelligence
adoption, AI analytics, and resilience metrics, provides the connection between conceptual theory and
managerial practice.
This research has limitations just like any other study. This study relies on secondary and publicly available data
sources. While this improves transparency and replicability, it limits access to firm-level proprietary operational
data and may not capture managerial decision constraints in specific organizations.
This work should be expanded in a number of directions in future research. The resilience strategies during
cross-border settings where the interdependencies are even more complicated could be evaluated in cross-border
supply chain studies. The hybrid AI models with statistical forecasting and machine learning might be evaluated
in terms of further performance improvements. Lastly, sector specific research- such as in pharmaceuticals,
textiles, or aerospace- would offer industry specific information on resilience strategies to industries that are
critical to the U.S. economic competitiveness and security.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 411
www.rsisinternational.org
ACKNOWLEDGEMENTS / FUNDING
The author appreciates the fact that the Atlantis University, Miami has assisted her by offering access to
academic materials and research advice. This study did not receive any external funding. The author states that
no financial conflicts of interests are present which may have affected the results of this study.
Ethical Considerations
This study relied on secondary data and information that are publicly available. No human-subject data were
collected. Data processing and visualization were conducted using Python (Google Colab) and matplotlib.
Interpretation and conclusions were produced by the author based on the reported analytical outputs.
Data Availability
All secondary datasets used in this study are publicly available from the U.S. Census Bureau, U.S. Bureau of
Labor Statistics, and public logistics datasets cited in the References. All transformations and evaluation metrics
described in the Methodology can be replicated using the cited public datasets and standard forecasting libraries.
REFERENCES
1. A. Ali, A. Mahfouz, and A. Arisha, “Analysing supply chain resilience: integrating the constructs in a
concept mapping framework via a systematic literature review,” Supply Chain Management: An
International Journal, vol. 22, no. 1, pp. 16–39, 2017, doi:10.1108/SCM-06-2016-0197.
2. U. Jüttner, H. Peck, and M. Christopher, “Supply chain risk management: outlining an agenda for future
research,” International Journal of Logistics: Research and Applications, vol. 6, no. 4, pp. 197–210, 2003,
doi:10.1080/13675560310001627016.
3. C. S. Tang, “Robust strategies for mitigating supply chain disruptions,” International Journal of Logistics:
Research and Applications, vol. 9, no. 1, pp. 33–45, 2006, doi:10.1080/13675560500405584.
4. L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001,
doi:10.1023/A:1010933404324.
5. G. E. P. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time Series Analysis: Forecasting and
Control, 5th ed. Hoboken, NJ, USA: Wiley, 2015.
6. T.-M. Choi, “Big data analytics in operations management,” Production and Operations Management,
vol. 27, no. 10, pp. 1868–1883, 2018, doi:10.1111/poms.12838.
7. S. Chopra, Supply Chain Management: Strategy, Planning and Operation, 7th ed. Harlow, U.K.: Pearson
Education, 2020. (Global Edition, ISBN 1292294833).