
Purpose This study aims to develop a calibration-aware hybrid framework for large-scale business and financial risk classification, addressing heterogeneous features, nonlinear relationships, temporal distribution shifts, class imbalance and unreliable probability estimates. Design/methodology/approach The hybrid multi-agent quantum neural transformer (H–MAQNT) combines a tabular transformer, domain-specialized variational quantum neural network (VQNN) agents and adaptive attention-based fusion. It uses leakage-controlled preprocessing, domain-informed features, imbalance-aware training, validation-only model selection and calibration and quantum-aware multi-fidelity Bayesian optimization (QMFBO) for cost-aware tuning. Findings On temporally held-out LendingClub and IEEE–CIS future test periods, H–MAQNT outperforms strong gradient-boosted, neural and transformer-based baselines in discrimination and probability quality. It improves the receiver operating characteristic area under the curve (ROC–AUC) by approximately 1.1–1.3 percentage points over the strongest Transformer-based baselines, lowers expected calibration error (ECE) on both data sets and reduces high-fidelity optimization evaluations by 35% relative to single-fidelity Bayesian optimization. Research limitations/implications The evidence is based on two public data sets and classically simulated quantum circuits. Future work should test additional institutions, rolling time windows, physical hardware and prospective drift and recalibration policies. Practical implications Calibrated risk probabilities may support fraud-alert prioritization, risk ranking, threshold setting, and secondary review, helping institutions weigh predictive gains against computational cost. For LendingClub, the inclusion of loan grade and interest rate makes the framework more appropriate for post-pricing monitoring than for independent pre-pricing approval. Operational use would require institution-specific validation, monitoring for drift and subgroup performance, documented decision thresholds, and human oversight. Social implications The study integrates tabular attention, domain-specialized simulated quantum representations, calibration analysis and cost-aware optimization in one decision-support framework, while evaluating incremental value without claiming inherent quantum advantage. Originality/value The contribution is the theoretically motivated integration of tabular attention, domain-specialized simulated quantum representations, calibration analysis and cost-aware optimization in a unified financial decision-support framework. The study does not claim inherent quantum advantage; instead, it tests the incremental value of simulated quantum modules through temporal evaluation, strong nonhybrid comparisons and component-level ablations.
Purpose Does technology destabilize financial systems, or does it merely redistribute risk across them? This paper aims to contribute to this debate by introducing and empirically testing a novel transmission mechanism, named cross-tech contagion, through which technological similarity may facilitate the propagation of financial disruption. Design/methodology/approach Using an initial sample covering up to 135 countries between 2008 and 2024, the authors integrate binary panel models to explain systemic crises, graph-theoretic network analysis, as well as spatial panel econometrics to identify the existence and extent of this channel. Findings Across specifications confirm that an increase of one standard deviation in the level of technological readiness raises the average predicted probability of a systemic banking crisis by 1.5–1.8 percentage points, compared to a baseline occurrence rate of 4.5%. Similarly, moving from the first to the third quartile of readiness increases the predicted probability from 3.0% to 4.8%. Network analysis illustrates the existence of a two-speed systemic architecture: a systemic core of technologically advanced countries and a peripheral cluster of highly interconnected technologically emerging countries. In addition, spatial dependence models indicate that the indirect effect is more dominant than the direct one: the indirect effect of the cross-tech spillover of crises substantially exceeds the direct domestic effect, providing plausible evidence consistent with a technological contagion channel. Originality/value These results justify macroprudential regulation paying particular attention to potential vulnerabilities related to technology.
Purpose The purpose of this paper is to develop a novel two-stage data envelopment analysis (DEA) framework that addresses the critical gap between high-level financial targets and granular operational drivers. This research aims to provide a deterministic modeling approach to derive attainable and justifiable corporate performance targets. Design/methodology/approach Accounting identities are embedded into a two-stage DEA production framework to decompose a composite indicator into performance drivers, efficiency regulators (ER) and structural capacity drivers. Stage 1 identifies implementable input targets through radial contraction of ER and slack minimization. Stage 2 determines attainable output expansion and final target levels under variable returns to scale. Exogenous growth requirements (MGG) and efficiency guardrails (MGE) are incorporated as managerial parameters defining the feasible target region. The framework is applied to six Japanese automakers using FY2023 data. Findings The framework provides explicit benchmarking-based justification for allocating a composite-indicator target to decomposed financial components. The empirical results show that attainable target return on equity (ROE) as a function of the sales growth rate can exhibit a peak, implying the existence of an optimal growth rate that maximizes attainable target ROE. Higher growth does not necessarily increase attainable target ROE. The solution structure around the peak also helps clarify ROE improvement strategies, including the relative emphasis on growth versus efficiency improvement. Research limitations/implications The empirical analysis is limited to one industry and one fiscal year. Future research can extend the framework to multi-period settings, other industries and other composite financial indicators and examine robustness under alternative managerial parameter settings. Practical implications The framework improves the implement ability of numerical targets by linking accounting structure, benchmark peers and managerial conditions within an explicitly defined feasible set. It can support management control and governance by clarifying feasible improvement paths and the tradeoff between growth and efficiency. Originality/value This paper integrates accounting identities and exogenous managerial parameters into a two-stage DEA framework, moving beyond black-box benchmarking to produce attainable, decomposed and managerially interpretable targets.
Purpose This study aims to examine whether firm-level market indicators and lagged macroeconomic information can be integrated within an explainable predictive framework for corporate financial distress, with particular attention to temporal dynamics, model performance and interpretability. Design/methodology/approach This study uses an unbalanced panel of 15,664 firm-month observations from 99 large-cap US listed equity series covering July 2012–September 2025. Regularized logistic regression, Random Forest, HistGradientBoosting and XGBoost are evaluated using chronological training, validation and held-out testing. SHapley Additive exPlanations (SHAP) provide model interpretability, while alternative macroeconomic lag structures, calibration diagnostics and robustness checks assess the stability of the predictive evidence. Findings Regularized logistic regression records the strongest full-test point estimates (AUC = 0.7886; average precision = 0.3085), while XGBoost is the strongest nonlinear benchmark (AUC = 0.7314; average precision = 0.2049). However, model rankings vary across evaluation periods. The combined one-, three- and six-month lag specification performs best, indicating that predictive macroeconomic information is distributed across multiple horizons. SHAP identifies 63-day firm-level volatility as the leading XGBoost feature and reveals conditional relationships between macroeconomic conditions and firm-specific vulnerability. Research limitations/implications The large-cap US sample, market-based distress definition and mixed-frequency design limit broader generalization. Practical implications The framework can support risk ranking, early-warning screening and interpretable financial-risk monitoring for investors, financial institutions and regulators. Originality/value This study integrates temporal macroeconomic information, comparative model evaluation and explainable AI, demonstrating that greater model complexity does not necessarily improve out-of-sample distress prediction.
Purpose This paper systematically reviews the use of machine learning (ML) and data-driven models in financial risk and liquidity management within large-scale projects, targeting the multifaceted challenges arising from economic, contractual and contextual complexities. This study aims to explore how data-centric approaches can strengthen forecasting, control cash flow volatility and improve resilience across infrastructure, energy and public-sector projects. Design/methodology/approach Following the PRISMA protocol, 142 peer reviewed studies published between 2012 and 2025 were comprehensively screened and analyzed. The sample includes 40 % quantitative, 32 % mixed methods, 16 % literature review and 11 % qualitative designs, demonstrating a balanced methodological landscape. The studies were synthesized across three interrelated domains, i.e. (i) data, (ii) model and (iii) application to capture technical, analytical and contextual dimensions of financial risk management. Findings The reviewed literature reveals clear research concentration patterns, with ML-based models, particularly ensemble methods, neural networks and decision tree algorithms, emerging as the dominant analytical tools, especially in credit risk assessment and related financial decision contexts. Credit risk-related research (30%) dominates the literature, while project-level liquidity remains underexplored (≈ 4%), marking a significant conceptual gap. Persistent limitations include lack of accurate historical data (28%), model transparency issues (19%) and market-specific bias (18%). These constraints highlight the need to improve data quality, enable interpretability and adapt models across heterogeneous project contexts and economic systems. Research limitations/implications This review is limited by potential selection bias due to language restrictions and the exclusion of unpublished or gray literature. The heterogeneity in methodological quality and reporting among included studies complicates direct comparison and synthesis of findings. The reliance on secondary data and simulation models also underscores the need for more empirical, context-specific studies. Future research should aim to address these gaps, explore broader regional contexts and validate ML models under real-world project conditions. Practical implications The findings highlight the superior predictive power of ML models in managing financial risks and liquidity in large-scale projects, facilitating more informed and proactive decision-making. Practitioners can use hybrid and interpretable models to optimize risk assessment processes and adapt financial strategies to dynamic project environments. However, effective implementation requires addressing data quality issues, ensuring model adaptability to local conditions and enhancing stakeholder collaboration across project teams. Integrating advanced analytics into existing project finance frameworks can drive more resilient, data-driven risk management and improve overall project outcomes. Social implications Adopting advanced ML and data-driven risk management methods can enhance transparency and accountability in project finance, supporting socioeconomic development and stability. Improved risk mitigation reduces the likelihood of project failures, cost overruns and financial crises that impact broader communities. Better liquidity management enables timely delivery of essential infrastructure, public services and employment, contributing to sustainable urban growth and improved living standards. Bridging the digital divide and addressing ethical considerations in AI usage will ensure equitable benefits for diverse stakeholders and promote public trust in technological advancements. Originality/value By integrating findings within the three-domain (data–model–application) framework, this review offers a structured synthesis of ML’s impact on financial risk and liquidity management. It delineates empirical performance, contextual adaptability and methodological challenges providing actionable directions for bridging theoretical design and real-world decision systems. The study delivers a clear roadmap for researchers and policymakers to strengthen transparency, data governance and cross-disciplinary collaboration in project finance analytics. Rather than promoting specific algorithms, this review maps the structure, focus and blind spots of existing research, providing a consolidated evidence base for future methodological and application-oriented studies.