AI-Guided Diversification of Green Technology Supply Chains

Dmitry Erokhin

Abstract

Green technology supply chains combine rapid demand growth with concentrated production, uneven environmental performance, and exposure to trade disruption. This article links trade network analysis, next-year bilateral trade link prediction, multi-objective portfolio optimization, and disruption stress testing across 14 importing markets and six technology and material groups from 2015 to 2024. The cleaned sample contains 15,186 supplier-year observations and USD 696.1 billion in foreign supplier trade. Concentration rose most clearly in photovoltaic equipment, lithium-ion batteries, and wind equipment. In the chronological 2022–2023 test, gradient boosting achieved ROC AUC 0.911 and a Brier score of 0.119. Logistic regression produced slightly higher AUC, while persistence retained the strongest F1. Event-specific tests show that exits and reactivations are more predictable than rare first observed entries. A historical decision comparison finds only a small and statistically uncertain advantage of calibrated gradient boosting over calibrated logistic regression, while predictive continuity information improves allocation performance relative to a specification without the probability term. Latest-year portfolios remain highly concentrated, with median HHI of 0.632 and a median top-three supplier share of 95.5%. Under balanced priorities, median HHI falls by 53.7%, and sourcing exposure to supplier-country CO2 intensity of GDP falls by 32.3%. These changes are accompanied by a 5.2% median unit value increase and a 4.8 percentage point decline in predicted next-year link activity. Joint supplier budget scenarios show that individually feasible national diversification plans can compete for the same alternatives. The framework provides a transparent basis for screening sourcing options, identifying cases that require supplier-level verification, and assessing where coordinated investment may be needed.

Source: semanticscholar · PDF

Read the AI summary, key takeaways and discussion on WOBR Quant Research.