Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM67281
Volume:
10 |
9 |
Sep 2026
Pages:
1-14
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Engineering
Pillai College of Engineering
Computer Engineering
Pillai College of Engineering
Computer Engineering
Pillai College of Engineering
Computer Engineering
Pillai College of Engineering
Computer Engineering
Pillai College of Engineering
Machine learning is increasingly being used by financial institutions for approving loans, forecasting credit-risk, for underwriting insurance, and in fraud detection, among other applications, but the machine learning systems the institutions use are often opaque to the people whose lives they touch and affect: the loan applicant, the loan auditor and the financial supervisor. A consolidated survey of twenty recent studies on how to implement Explainable Artificial Intelligence (XAI) in financial decision making is presented, grouped into four research themes: loan approval and financial/credit risk assessment, insurance and personalization, financial and insurance fraud detection, and know your customer (KYC) identity verification and financial-document intelligence. The survey is able to benchmark the algorithms including machine learning models, explainability techniques, data sets, and the results reported over the literature, but does not suggest a "one fits all" approach to comparing performance across different studies. Even though every pipeline stage -identity verification, document processing, fraud screening, predictive decisioning, post-hoc explanation- is well represented on its own, nothing from the literature has been found of an integrated system that underpins all these stages in a user-centric workflow for the benefit of the applicant. Given this disintegration, the paper proposes a data-consent based all-in-one explainable artificial intelligence system for loans and insurance decision making, which consists of a combination of biometric KYC verification, OCR-based document intelligence, anomaly detection based on fraud screening, gradient boosting based on risk prediction, explanation based on SHAP and a data access layer behind a consent. ArthaX places itself throughout as a research direction, not an experimentally proven system, and the survey concludes with identifying the fairness, explanation-stability, dataset-imbalance, regulatory and privacy challenges that remain open for trustworthy and end-to-end explainable financial decision systems.
KEYWORDS
This survey has analyzed twenty recent papers pertaining to explainable AI in loan approval, credit-risk assessment, insurance underwriting, fraud detection, KYC, and financial-document intelligence, summarizing their models, XAI techniques, data, and results. The evidence suggests that the most dominant and well-validated method for combining tree models with explainability is gradient-boosting and ensemble tree models; the explainability research of insurance, fraud, KYC, and document-intelligence is more fragmented and less consistently associated with tree models. Importantly, no reviewed study brings together identity verification, document processing, fraud screening, predictive decisioning and explanation to the applicant in an end-to-end system. This gap inspires ArthaX, an integrated explainable AI (XAI) direction made up of six modules: FaceNet-based KYC, OCR-based document intelligence, Isolation Forest-based fraud detection, XGBoost-based prediction, SHAP-based explanation, personalized recommendation, and Account Aggregator-based data access. The presented approach of ArthaX is not empirical but rather a research direction based on reviewed literature and future research. More generally, this survey indicates that the future of financial XAI research should shift from post hoc and post financial towards integrated and financial applicant-centered and verified consistent explainable decision systems.
[1] C. -A. Wilson, "Explainable AI in Finance: Addressing the Needs of Diverse Stakeholders," CFA Institute Research and Policy Center, 2025. [2] R. Akter, R. K. Halder, M. Rahman, Md. A. Uddin, A. Khraisat, and M. N. Uddin, "Prediction with SHAP-guided feature selection and LIME-based explainability for loan approval," International Journal of Intelligent Systems, 2026. [3] L.-H. Li, A. K. Sharma, and S.-T. Cheng, "Explainable AI-based LightGBM prediction model for default borrowers on social lending platforms," Intelligent Systems with Applications, vol. 26, Art. no. 200514, 2025. [4] V. Ravi, S. R. Mani Sekhar, and S. Siddesh, "Explainable AI (XAI) for credit scoring and loan approvals," Journal of Financial Data Science, 2025. [5] M. N. Sowmiya, Jaya Sri S., Deepshikha S., and Hanushya Devi G., "Credit risk analysis using explainable artificial intelligence," Journal of Soft Computing Paradigm, vol. 6, no. 3, 2024. [6] M. K. Nag et al., "Credit risk assessment and financial decision support using XAI," Risks (MDPI), vol. 12, Art. no. 164, 2024. [7] S. Kakkar, "Explainable AI models for credit risk scoring in banking: Balancing accuracy and regulatory transparency," International Journal of Financial Data Science (IJFDS), vol. 3, no. 2, pp. 1–6, 2025. [8] K. Madhu, V. Shirisha, A. Sonya, and L. R. Chandra, "Future of loan approvals using explainable AI," International Journal of Scientific Research in Engineering Trends (IJSRET), vol. 11, no. 3, 2025. [9] C. S. Kubam, "Agentic AI for autonomous, explainable, and real-time credit risk decision-making," International Journal of Intelligent Systems and Applications in Engineering (IJISAE), vol. 12, no. 23s, pp. 3669–3676, 2024. [10] Y. Shi, Z. Yang, Y. Wang, and L. Zhou, "CreditXAI: A multi-agent system for explainable corporate credit rating," arXiv:2510.22222, 2025.
Publication Date
9/15/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
1
Total views since publication
PDF Downloads
4
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...