804 résultats
pour « Autre »
This study presents the Bayesian Component Encoder Analysis (BCEA) method for identifying financial crime. Integrating FinBERT, principal component analysis (PCA), and Bayesian networks, BCEA uses financial texts, images, and transaction records. FinBERT extracts semantic features, PCA reduces data complexity, and Bayesian networks model features for probabilistic reasoning. The authors claim BCEA achieves 94.35% accuracy and a 12.78-second recognition time, surpassing LSTM and BERT models. The authors state that the method demonstrates potential for financial supervision and risk management, with possible applications in complex financial scenarios, based on experimental results validating its effectiveness.
This paper analyzes a bivariate optimal dividend problem for an insurer with two collaborating business lines under a diffusion model with correlated Brownian motions. The framework incorporates dividend payouts, proportional reinsurance, and inter-line capital transfers to prevent bankruptcy. The authors provide complete analytical solutions, identifying three scenarios with closed-form value functions and optimal strategies. Results show a threshold dividend policy, with the more important line having a lower threshold. Optimal reinsurance decreases with aggregate reserves and stabilizes after a switching point. Correlation between lines affects reinsurance, and the capital transfer rule is consistent across scenarios.
𝗘𝗜𝗢𝗣𝗔 released its July 2025 𝙄𝙣𝙨𝙪𝙧𝙖𝙣𝙘𝙚 𝙍𝙞𝙨𝙠 𝘿𝙖𝙨𝙝𝙗𝙤𝙖𝙧𝙙, offering an assessment of the European insurance sector's financial health as of Q1 2025 Solvency II data and Q2 2025 market data. Overall, the report indicates a stable risk landscape at a medium level for the European insurance sector, demonstrating notable resilience. However, it also highlights a negative outlook in certain areas over the next year, influenced by complex global dynamics such as geopolitical tensions and market volatility. Specifically, market risks due to fixed income volatility and cyber and digitalization risks are identified as growing concerns, necessitating continued vigilance despite general stability.
This paper introduces an 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝘃𝗲 𝗵𝘆𝗯𝗿𝗶𝗱 𝗶𝗻𝘀𝘂𝗿𝗮𝗻𝗰𝗲 𝗺𝗼𝗱𝗲𝗹 designed to cover 𝗵𝗲𝗮𝘃𝘆-𝘁𝗮𝗶𝗹𝗲𝗱 𝗹𝗼𝘀𝘀𝗲𝘀, which are extreme and potentially limitless financial damages, often associated with natural disasters. 𝗧𝗵𝗲 𝗺𝗼𝗱𝗲𝗹 𝗰𝗼𝗺𝗯𝗶𝗻𝗲𝘀 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗶𝗻𝗱𝗲𝗺𝗻𝗶𝘁𝘆-𝗯𝗮𝘀𝗲𝗱 𝗶𝗻𝘀𝘂𝗿𝗮𝗻𝗰𝗲 𝗳𝗼𝗿 𝘀𝗺𝗮𝗹𝗹𝗲𝗿 𝗹𝗼𝘀𝘀𝗲𝘀 𝘄𝗶𝘁𝗵 𝗽𝗮𝗿𝗮𝗺𝗲𝘁𝗿𝗶𝗰 (𝗶𝗻𝗱𝗲𝘅-𝗯𝗮𝘀𝗲𝗱) 𝗶𝗻𝘀𝘂𝗿𝗮𝗻𝗰𝗲 𝗳𝗼𝗿 𝗹𝗮𝗿𝗴𝗲𝗿, 𝗰𝗮𝘁𝗮𝘀𝘁𝗿𝗼𝗽𝗵𝗶𝗰 𝗲𝘃𝗲𝗻𝘁𝘀. A key contribution is the development of a 𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗲𝗱 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗰𝗿𝗶𝘁𝗲𝗿𝗶𝗼𝗻 and a 𝘁𝘄𝗼-𝘀𝘁𝗲𝗽 𝗰𝗮𝗹𝗶𝗯𝗿𝗮𝘁𝗶𝗼𝗻 𝗺𝗲𝘁𝗵𝗼𝗱𝗼𝗹𝗼𝗴𝘆 that can leverage readily available covariate data, even when comprehensive loss data is scarce. Empirical analysis using both 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗲𝗱 𝗮𝗻𝗱 𝗿𝗲𝗮𝗹-𝘄𝗼𝗿𝗹𝗱 𝘁𝗼𝗿𝗻𝗮𝗱𝗼 𝗱𝗮𝘁𝗮 demonstrates that 𝘁𝗵𝗶𝘀 𝗵𝘆𝗯𝗿𝗶𝗱 𝗰𝗼𝗻𝘁𝗿𝗮𝗰𝘁 𝗼𝘂𝘁𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝘀 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗮𝗽𝗽𝗲𝗱 𝗶𝗻𝗱𝗲𝗺𝗻𝗶𝘁𝘆 𝗰𝗼𝗻𝘁𝗿𝗮𝗰𝘁𝘀 by providing better coverage for the same premium, especially benefiting regions with limited data. The authors highlight the practical advantages of 𝗳𝗮𝘀𝘁𝗲𝗿 𝗰𝗼𝗺𝗽𝗲𝗻𝘀𝗮𝘁𝗶𝗼𝗻 and 𝗿𝗲𝗱𝘂𝗰𝗲𝗱 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗼𝘀𝘁𝘀 offered by the parametric component.
This academic paper proposes these 𝗸𝗲𝘆 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀:
• The analysis provides a framework for introducing index insurance in competition with traditional products, emphasizing demand and solvency.
• Key drivers for index insurance demand are policyholder risk aversion, compensation speed advantage over traditional products, and its pricing (loading factor).
• The proposed hybrid product effectively balances the strengths of both insurance types by applying index insurance where it is “most suitable for policyholders,” accelerating compensation, and potentially reducing premiums.
• The methodology can help insurers identify specific loss types for which index compensation is preferred, optimizing portfolio structure and claims management.
• Future work will address modeling demand for index insurance in situations where traditional indemnity-based insurance is unavailable, requiring a “more nuanced approach to calibrate the utility function.”
This document introduces a novel two-step methodology for money laundering detection that significantly improves upon existing rule-based and traditional machine learning methods. The first step involves representation learning using a transformer neural network, which analyzes complex financial time series data without requiring labels through contrastive learning. This self-supervised pre-training helps the model understand the inherent patterns in transactions. The second step then leverages these learned representations within a two-threshold classification procedure, calibrated by the Benjamini-Hochberg (BH) procedure, to control the false positive rate while accurately identifying both fraudulent and non-fraudulent accounts, addressing the significant class imbalance in money laundering datasets. Experimental results on real-world, anonymized financial data demonstrate that this transformer-based approach outperforms other models in detecting fraudulent activities.
This consultation package is aimed at easing the reporting burden on insurance and reinsurance companies under the Solvency II framework. The proposed amendments seek to reduce reporting requirements by at least 26% for solo undertakings and 36% for small and non-complex undertakings. Key changes include reducing template frequency, deleting annual templates, and introducing technical simplifications. The EIOPA expects these changes to substantially reduce the burden on European insurers without compromising policyholder protection or financial stability. Stakeholders can provide feedback via the EU Survey until October 10, 2025.
This study develops a machine learning framework to identify high-risk enterprise financial reports, comparing Support Vector Machine, Random Forest, and K-Nearest Neighbors models. Using 2020–2025 audit data from the Big Four firms, Random Forest showed the highest performance (F1-score: 0.9012), excelling in detecting fraud and compliance issues. While KNN struggled with high-dimensional data, SVM performed well but was computationally intensive. The study highlights the potential of machine learning in auditing but notes limitations, including reliance on structured data and exclusion of external economic factors.
En 2024, Tracfin a franchi le cap des 200 000 déclarations de soupçon, avec 211 165 signalements (+13,2 % par rapport à 2023), reflétant l’engagement croissant des 50 professions assujetties à la lutte contre le blanchiment de capitaux (LCB-FT). Le secteur financier domine (93,1 %), mais le non-financier progresse (+25,7 %), notamment les opérateurs d’art (+254,4 %). Deux nouvelles professions, les entreprises de jeux numériques et gestionnaires de crédit, intègrent le dispositif. Tracfin renforce la qualité des déclarations via des échanges avec les déclarants et consolide sa coopération internationale, notamment avec l’AMLA et le Groupe Egmont.
This report examines how European (re)insurers address biodiversity risks, which threaten financial stability due to their complexity and links with climate risks. Despite challenges in quantifying impacts, one in five insurers references biodiversity in their risk assessments, though mostly qualitatively. Promising practices show growing awareness, but regional variations and limited metrics hinder progress. EIOPA calls for enhanced collaboration to improve data, models, and risk management, emphasizing the need to better understand the climate-biodiversity nexus and explore nature-based solutions to address insurance gaps.