Adaptive Management of Commercial Banks' Securities Portfolios Using Explainable Artificial Intelligence (Educational Yield Insights & Breakthroughs)

Summary

• The share of government and Central Bank instruments in the portfolio is above 96%, up from above 90%, showing concentration in a single asset type.p.1
• Under the baseline scenario the portfolio-to-equity ratio is forecast to exceed 100 percent in 2029–2030.p.1
• According to the Bank of England and FCA, 75 percent of financial firms already use artificial intelligence and 46 percent acknowledge only a partial understanding of the technologies they use.p.2
• The empirical basis is panel and ARDL models on panel data for 30 banks (417 observations) and monthly data for five banks over 2020–2025.p.2
• The proposed XAI-DPO methodology combines machine learning forecasts, SHAP explanations and dynamic optimization in a closed loop of five blocks.p.3

Key insights

1. Portfolio yield is 9.0% against a return on assets of 15.1%, which indicates the portfolio's income potential is underused.p.1
2. 81 percent of firms using artificial intelligence apply explainability methods and 64 percent use the SHAP method.p.2
3. BlackRock's Aladdin platform analyzes portfolio risk for more than 190 institutions, including central banks.p.2
4. Presidential Resolution No. PQ-358 of 14 October 2024 approved the Strategy for the Development of Artificial Intelligence Technologies until 2030.p.2
5. The liquidity coverage ratio is the strongest positive factor and the exchange rate the strongest negative one at banking-system level, but the effects vary sharply across banks.p.2
6. In a stress regime, with FX factor contribution above 35% and VaR limit utilization above 90%, the risk coefficient is 2× baseline and the portfolio is reviewed weekly.p.4
7. Each rebalancing decision is submitted to the Asset and Liability Management Committee together with an explanation report, preserving human oversight.p.4