Risks

Research

33 pages, 2622 KB

Open AccessFeature PaperArticle

Enhancing Enterprise Risk Management and Internal Audit Practices by Applying Machine Learning Models

by Reneta Duhova, Angel Duhov, Petia Georgieva and Milena Lazarova

Risks 2026, 14(5), 107; https://doi.org/10.3390/risks14050107 - 6 May 2026

Viewed by 312

Organizations are currently in a stage where the volume of financial transactions and data is constantly growing. The same goes for risks associated with the use of data for risk management and strategic decision-making. The likelihood of transactional errors generally increases with data [...] Read more.

Organizations are currently in a stage where the volume of financial transactions and data is constantly growing. The same goes for risks associated with the use of data for risk management and strategic decision-making. The likelihood of transactional errors generally increases with data volume and process complexity, while fraud, although less frequent, may have more severe financial, compliance, and reputational consequences for organizations. Continuous auditing practices and well-established enterprise risk management (ERM) processes, combined with AI-driven pattern recognition, trend analysis and segmentation, can enhance timely detection and proper investigation of suspicious transactions. In areas with large volumes of transactions, the audit sampling process may be a lengthy process and pose a detection risk. Using machine learning (ML) models to support critical business processes could prove effective in managing enterprise risk overall. The current study offers new perspectives on managing risk and assurance with ML model output for flagging possible risky transactions within ERP (SAP) systems data. The study population consists of 69,158 finalized billing records extracted from the SAP production environment of a private sector organization, which covers a six-month operational period. The dataset was divided into an 80/20 train–test split, yielding 55,326 training and 13,832 test instances across six classification categories. The study examines the ML methods’ outcomes from billing datasets and their applicability in enhancing audit, assurance, and ERM processes by evaluating output data results from two supervised classification algorithms—multinomial logistic regression (SoftMax regression) and XGBoost—against various criteria generally accepted as risky in audit engagements. Model performance was assessed using accuracy, precision, recall, F1-score, ROC-AUC, and average precision (AP) from precision–recall curves. The results confirm that XGBoost achieves 99% overall accuracy with a macro F1-score of 0.965, outperforming logistic regression (macro F1 = 0.863), and that ML output allows early investigation and follow-up procedures to minimize the risk of fraud and errors and optimize risk management activities, thus strengthening internal control frameworks. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

24 pages, 921 KB

Open AccessFeature PaperArticle

Advanced Insurance Risk Modeling for Pseudo-New Customers Using Balanced Ensembles and Transformer Architectures

by Finn L. Solly, Raquel Soriano-Gonzalez, Angel A. Juan and Antoni Guerrero

Risks 2026, 14(4), 91; https://doi.org/10.3390/risks14040091 - 17 Apr 2026

Viewed by 633

Abstract

In insurance portfolios, classifying customers without a prior history at a given company is particularly challenging due to the absence of historical behavior, extreme class imbalance, heavy-tailed loss distributions, and strict operational constraints. Traditional machine learning approaches, including the baseline methodology proposed in [...] Read more.

In insurance portfolios, classifying customers without a prior history at a given company is particularly challenging due to the absence of historical behavior, extreme class imbalance, heavy-tailed loss distributions, and strict operational constraints. Traditional machine learning approaches, including the baseline methodology proposed in previous studies, typically optimize global predictive accuracy and therefore fail to capture business-critical outcomes, especially the identification of high-risk clients. This study extends the existing approach by evaluating two complementary business-aware classification strategies: (i) a balanced bagging ensemble specifically designed to handle class imbalance and maximize expected profit under explicit customer-omission constraints, and (ii) a lightweight Transformer-based architecture capable of learning richer feature representations. Both approaches incorporate the asymmetric financial cost structure of insurance and operate under operational selection limits. The empirical analysis is conducted on a proprietary large-scale auto insurance dataset comprising 51,618 customers and is complemented by validation on nine synthetic datasets to assess robustness. Model performance is evaluated using statistical tests (ANOVA, Friedman, and pair-wise comparisons) together with business-oriented metrics. The results show that both proposed approaches consistently outperform the baseline methodology (p < 0.001) in terms of profit, with the ensemble offering a better balance of performance and efficiency, while the Transformer shows stronger robustness and generalization under data perturbations. The balanced ensemble provides the most favourable trade-off between predictive performance, robustness, interpretability, and computational efficiency, making it suitable for deployment in regulated insurance environments, while the Transformer achieves competitive results and exhibits stronger generalization under data perturbations. The proposed approach aligns machine learning with actuarial portfolio optimization by explicitly integrating profit-driven objectives and operational constraints, offering two practical and scalable solutions for risk-based decision-making in real-world insurance settings. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

14 pages, 1488 KB

Open AccessFeature PaperArticle

A Framework for Interpreting Machine Learning Models in Bond Default Risk Prediction Using LIME and SHAP

by Yan Zhang, Lin Chen and Yixiang Tian

Risks 2026, 14(2), 23; https://doi.org/10.3390/risks14020023 - 28 Jan 2026

Cited by 2 | Viewed by 1428

Abstract

Interpretability analysis methods, such as LIME and SHAP, are widely employed to explain the predictions of artificial intelligence models; however, they primarily function as post hoc tools and do not directly quantify the intrinsic interpretability of the models. Although it is commonly assumed [...] Read more.

Interpretability analysis methods, such as LIME and SHAP, are widely employed to explain the predictions of artificial intelligence models; however, they primarily function as post hoc tools and do not directly quantify the intrinsic interpretability of the models. Although it is commonly assumed that model transparency decreases with increasing complexity, there is currently no standardized framework for evaluating interpretability as an inherent property of AI models. In this study, we examine the prediction of bond defaults using several widely used machine learning algorithms. The classification performance of each algorithm is first evaluated, followed by the application of LIME and SHAP to assess the influence of input features on model outputs. Based on these analyses, we propose a novel approach for quantifying intrinsic model interpretability. The results align with theoretical expectations and provide insights into the trade-off between model complexity and interpretability. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

31 pages, 2487 KB

Open AccessArticle

Enhancing Predictive Performance of LSTM–Attention Models for Investment Risk Forecasting

by Amina Ladhari and Heni Boubaker

Risks 2026, 14(1), 13; https://doi.org/10.3390/risks14010013 - 5 Jan 2026

Cited by 1 | Viewed by 1590

Abstract

For many decades, time-series forecasting has been applied to different problems by scientists and industries. Many models have been introduced for the purpose of forecasting. These advancements have significantly improved the accuracy and reliability of predictions, especially in complex scenarios where traditional methods [...] Read more.

For many decades, time-series forecasting has been applied to different problems by scientists and industries. Many models have been introduced for the purpose of forecasting. These advancements have significantly improved the accuracy and reliability of predictions, especially in complex scenarios where traditional methods struggled. As data availability continues to expand, the integration of machine learning techniques is likely to further enhance forecasting capabilities across various fields. Today, hybrid techniques are gaining popularity, as they combine the advantages of different approaches to deliver improved predictive performance and more advanced visualization analytics for decision support. These hybrid approaches can provide better prediction, and at the same time, they can develop a more sophisticated set of visualization analytics for decision support. Recently, the integration of cross-entropy, fuzzy logic, and attention mechanisms in hybrid forecasting models has enhanced their ability to capture complex and uncertain patterns in financial and energy markets. In this study, we propose a hybrid ANN–LSTM deep learning model optimized with cross-entropy, fuzzy logic, and an attention mechanism to enhance the forecasting of financial and energy time series, specifically Ethereum and natural gas prices. Our models combine the feature extraction strength of ANN with the temporal learning of LSTM, while cross-entropy improves convergence, fuzzy logic handles uncertainty, and attention refines feature weighting. Since inaccurate forecasts can lead to greater estimation uncertainty and increased financial and operational risk, improving predictive reliability is essential for effective risk mitigation. These techniques prove effective not only in improving estimation accuracy but also in minimizing financial risks and supporting more informed investment decisions. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

27 pages, 2446 KB

Open AccessArticle

Machine Learning & Artificial Intelligence Powered Credit Scoring Models for Islamic Microfinance Institutions: A Blockchain Approach

by Mohammad Mushfiqul Haque Mukit, Fakhrul Hasan, Tonmoy Choudhury, Amer Al Fadli and Abubaker Fadul

Risks 2026, 14(1), 12; https://doi.org/10.3390/risks14010012 - 5 Jan 2026

Cited by 2 | Viewed by 2710

Abstract

Islamic Microfinance Institutions (IMFIs) encounter distinct difficulties with credit scoring because they need to follow Shariah principles that combine riba bans with fair financial dealings regulations. Conventional credit scoring models exhibit two shortcomings: a poor capability to incorporate non-financial behavioral data and inadequate [...] Read more.

Islamic Microfinance Institutions (IMFIs) encounter distinct difficulties with credit scoring because they need to follow Shariah principles that combine riba bans with fair financial dealings regulations. Conventional credit scoring models exhibit two shortcomings: a poor capability to incorporate non-financial behavioral data and inadequate support for Islamic Microfinance Institutions’ requirements. Researchers use machine learning coupled with blockchain technology to create an adaptive Shariah-compliant credit scoring method that solves problems found in standard evaluation systems. Using a dataset of 1275 farmers with 52 weeks of transaction data, we implemented and compared three ML models: Linear Regression, Random Forest, and Gradient Boosting. Data preparation involved addressing 53% missing transaction data, followed by summing weekly financial activity to prepare it for predictive evaluations. Our analysis shows that the Random Forest model produced the best results with an R-squared value of 0.87 and a Mean Squared Error (MSE) of 12.4. In creditworthiness binary classification tasks, Gradient Boosting delivered an F1 score of 0.91 while maintaining precision at 0.89 and recall at 0.93. Blockchain integration exists to protect data through secure mechanisms that also conserve Islamic financial integrity and promote transparency. The research shows how ML and Blockchain technology enable fundamental changes in IMFIs by delivering elevated predictive accuracy, operational enhancements, and complete transparency. The conceptual framework guides ethical financial inclusion strategy by offering a solution for marginalized communities, but remains consistent with global sustainability objectives. The research established foundational elements for implementing cutting-edge technologies within IMFIs, which will promote new economic growth and build confidence in Shariah-compliant financial systems. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

19 pages, 1317 KB

Open AccessArticle

Metaheuristics for Portfolio Optimization: Application of NSGAII, SPEA2, and PSO Algorithms

by Ameni Ben Hadj Abdallah, Rihab Bedoui and Heni Boubaker

Risks 2025, 13(11), 227; https://doi.org/10.3390/risks13110227 - 19 Nov 2025

Cited by 1 | Viewed by 1356

Abstract

This work looks for the optimal allocation of different assets, namely, the G7 stock indices, commodities (gold and WTI crude oil), cryptocurrencies (Bitcoin and Ripple), and S&P Green Bond, over four periods: before the COVID-19 crisis, during the COVID-19 crisis and before the [...] Read more.

This work looks for the optimal allocation of different assets, namely, the G7 stock indices, commodities (gold and WTI crude oil), cryptocurrencies (Bitcoin and Ripple), and S&P Green Bond, over four periods: before the COVID-19 crisis, during the COVID-19 crisis and before the Russia–Ukraine war, during the COVID-19 crisis and Russia–Ukraine war, and after the COVID-19 pandemic and during the Russia–Ukraine war. Metaheuristics, Non-dominated Sorting Genetic Algorithm (NSGAII), Strength Pareto Evolutionary Algorithm (SPEA2), and Particle Swarm Optimization (PSO) are applied to find the best allocation. The results reveal that there a significant preference for the S&P Green Bond during the four periods of study according to three algorithms, thanks to its portfolio diversification abilities. During the COVID-19 pandemic and the geopolitical crisis, the most optimal portfolio was Nikkei 225 because of its quick recovery from the pandemic and poor reliance on the Russia–Ukraine markets, while WTI crude oil and both dirty and clean cryptocurrencies were poor contributors to the investment portfolio because these assets are sensitive to geopolitical problems. After the end of the pandemic and during the ongoing Russia–Ukraine war, the three algorithms obtained remarkably different results: the NSGAII portfolio was invested in various assets, 32% of the SPEA2 portfolio was allocated to the S&P Green Bond, and half of the PSO portfolio was allocated to the S&P Green Bond too. This may be due to changes in investors’ preferences to protect their fortune and to diversify their portfolio during the war. From a risk-averse perspective, NSGAII does not underestimate the risk, while in terms of forecasting accuracy, PSO is an adequate algorithm. In terms of time, NSGAII is the fastest algorithm, while SPEA2 requires more time than the NSGAII and PSO algorithms. Our results have important implications for both investors and risk managers in terms of portfolio and risk management decisions, and they highlight the factors that influence investment choices during health and geopolitical crises. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

25 pages, 1453 KB

Open AccessArticle

Application of Standard Machine Learning Models for Medicare Fraud Detection with Imbalanced Data

by Dorsa Farahmandazad, Kasra Danesh and Hossein Fazel Najaf Abadi

Risks 2025, 13(10), 198; https://doi.org/10.3390/risks13100198 - 13 Oct 2025

Cited by 1 | Viewed by 1939

Abstract

Medicare fraud poses a substantial challenge to healthcare systems, resulting in significant financial losses and undermining the quality of care provided to legitimate beneficiaries. This study investigates the use of machine learning (ML) to enhance Medicare fraud detection, addressing key challenges such as [...] Read more.

Medicare fraud poses a substantial challenge to healthcare systems, resulting in significant financial losses and undermining the quality of care provided to legitimate beneficiaries. This study investigates the use of machine learning (ML) to enhance Medicare fraud detection, addressing key challenges such as class imbalance, high-dimensional data, and evolving fraud patterns. A dataset comprising inpatient claims, outpatient claims, and beneficiary details was used to train and evaluate five ML models: Random Forest, KNN, LDA, Decision Tree, and AdaBoost. Data preprocessing techniques included resampling SMOTE method to address the class imbalance, feature selection for dimensionality reduction, and aggregation of diagnostic and procedural codes. Random Forest emerged as the best-performing model, achieving a training accuracy of 99.2% and validation accuracy of 98.8%, and F1-score (98.4%). The Decision Tree also performed well, achieving a validation accuracy of 96.3%. KNN and AdaBoost demonstrated moderate performance, with validation accuracies of 79.2% and 81.1%, respectively, while LDA struggled with a validation accuracy of 63.3% and a low recall of 16.6%. The results highlight the importance of advanced resampling techniques, feature engineering, and adaptive learning in detecting Medicare fraud effectively. This study underscores the potential of machine learning in addressing the complexities of fraud detection. Future work should explore explainable AI and hybrid models to improve interpretability and performance, ensuring scalable and reliable fraud detection systems that protect healthcare resources and beneficiaries. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

15 pages, 2189 KB

Open AccessArticle

AI Risk Management: A Bibliometric Analysis

by Adelaide Emma Bernardelli and Paolo Giudici

Risks 2025, 13(7), 131; https://doi.org/10.3390/risks13070131 - 7 Jul 2025

Cited by 5 | Viewed by 4616

Abstract

The growth of Artificial Intelligence applications requires the development of risk management models that can balance opportunities with risks. This paper contributes to the development of Artificial Intelligence risk management models by means of a thorough bibliometric analysis. The analysis highlights the need [...] Read more.

The growth of Artificial Intelligence applications requires the development of risk management models that can balance opportunities with risks. This paper contributes to the development of Artificial Intelligence risk management models by means of a thorough bibliometric analysis. The analysis highlights the need to develop a quantitative AI risk management framework. Full article

(This article belongs to the Special Issue Artificial Intelligence Risk Management)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Artificial Intelligence Risk Management

Share This Special Issue

Special Issue Editor

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Published Papers (8 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI