A Proxy-Information Bottleneck Strategy for Cognitive Credit Risk Detection.

Saved in:

Bibliographic Details
Title:	A Proxy-Information Bottleneck Strategy for Cognitive Credit Risk Detection.
Authors:	Chiappino, Simone¹ chiappino.simone@gmail.com
Source:	Engineering Letters. Feb2026, Vol. 34 Issue 2, p576-590. 15p.
Subjects:	Artificial intelligence, Credit analysis, Real-time computing, Financial services industry, Anomaly detection (Computer security), Dynamical systems, Feature selection
Abstract:	The digital transformation of financial services is rapidly increasing the volume and complexity of data that banks must process to deliver personalized and secure digital experiences. As banking services become increasingly digitized and remote, bridging the gap between trustworthy client information and effective fraud prevention remains a critical challenge. While artificial intelligence (AI) has proven effective in datadriven analytics, insights from cognitive science suggest that anomaly detection--including credit risk--can benefit from adaptive, context-aware mechanisms inspired by human cognition. Building on the Cognitive Dynamic Systems (CDS) framework of Haykin and Fuster, this paper introduces a Cognitive Node (CN) as a core element of a broader bio-inspired architecture for intelligent and trust-oriented financial ecosystems. The proposed model integrates a proxy Information Bottleneck (IB) principle within the CN (referred to as IB-CN), enabling the dynamic selection of the most relevant representations for credit risk detection. By optimizing an IB-driven objective, the system balances the trade-off between feature informativeness and predictive efficiency, thereby enhancing the identification of salient data. Experimental results show that the IB-CN improves truepositive detection while reducing false alarms in real-world credit risk scenarios. The model selectively discards noninformative features without sacrificing accuracy and consistently outperforms state-of-the-art machine learning baselines. These properties make the IB-CN particularly suitable for realtime credit-risk applications that require adaptive, contextsensitive decision-making and trust-aware risk strategies. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	The digital transformation of financial services is rapidly increasing the volume and complexity of data that banks must process to deliver personalized and secure digital experiences. As banking services become increasingly digitized and remote, bridging the gap between trustworthy client information and effective fraud prevention remains a critical challenge. While artificial intelligence (AI) has proven effective in datadriven analytics, insights from cognitive science suggest that anomaly detection--including credit risk--can benefit from adaptive, context-aware mechanisms inspired by human cognition. Building on the Cognitive Dynamic Systems (CDS) framework of Haykin and Fuster, this paper introduces a Cognitive Node (CN) as a core element of a broader bio-inspired architecture for intelligent and trust-oriented financial ecosystems. The proposed model integrates a proxy Information Bottleneck (IB) principle within the CN (referred to as IB-CN), enabling the dynamic selection of the most relevant representations for credit risk detection. By optimizing an IB-driven objective, the system balances the trade-off between feature informativeness and predictive efficiency, thereby enhancing the identification of salient data. Experimental results show that the IB-CN improves truepositive detection while reducing false alarms in real-world credit risk scenarios. The model selectively discards noninformative features without sacrificing accuracy and consistently outperforms state-of-the-art machine learning baselines. These properties make the IB-CN particularly suitable for realtime credit-risk applications that require adaptive, contextsensitive decision-making and trust-aware risk strategies. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X