[论文解读] AI-based Personalization and Trust in Digital Finance

Vijaya Kanaparthi|arXiv (Cornell University)|Jan 28, 2024

Impact of AI and Big Data on Business and Society被引用 20

一句话总结

本文对数字金融中以AI驱动的个性化与信任进行了系统性文献综述，并提出了一种基于AI的信用风险检测模型，使用四种分类器，具有竞争力的性能。

ABSTRACT

Personalized services bridge the gap between a financial institution and its customers and are built on trust. The more we trust the product, the keener we are to disclose our personal information in order to receive a highly personalized service that maximizes consumer value. Artificial Intelligence (AI) can help financial institutions tailor relevant products and services to their customers as well as improve their credit risk management, compliance, and fraud detection capabilities by incorporating chatbots and face recognition systems. The present study has analyzed sixteen research papers using the PRISMA model to perform a Systematic Literature Review (SLR). It has identified five research gaps and corresponding questions to analyze the present scenario. One of the gaps is credit risk detection for improved personalization and trust. Finally, an AI-based credit risk detection model has been built using four supervised machine learning classifiers viz., Support Vector Machine, Random Forest, Decision Tree, and Logistic Regression. Performance comparison shows an optimal performance of the model giving accuracy of ~89%, precision of ~88%, recall of ~89%, specificity of ~89%, F1_score of ~88%, and AUC of 0.77 for the Random Forest classifier. This model is foreseen to be most suitable for envisaging customer characteristics for which personalized credit risk mitigation strategies are particularly effective as compared to other existing works presented in this study.

研究动机与目标

• 通过信任与消费者价值来推动数字金融中的个性化。
• 综合现有关于个性化、风险与合规中的AI应用文献。
• 识别研究空白并提出引导未来工作的问题。
• 开发一个基于AI的信用风险检测模型，以提升个性化融资决策。

提出的方法

• 基于PRISMA的系统性文献综述，涵盖16篇论文。
• 识别空白并提出五个研究问题。
• 使用SVM、RF、DT和LR构建并评估信用风险检测模型。
• 报告性能指标，包括准确率、精确率、召回率、特异性、F1和AUC。

实验结果

研究问题

• RQ1 在数字金融中，基于AI的个性化与信任当前存在哪些空缺？
• RQ2 AI如何改进信用风险检测以支持个性化，同时维持信任？
• RQ3 哪些AI模型和特征在个性化风险缓释方面最有效？
• RQ4 基于AI的个性化在伦理、监管与合规方面有哪些考量？
• RQ5 在为个性化融资策略预测信用风险方面，不同分类器的比较如何？

主要发现

• 系统性文献回顾识别了与个性化、信任和信用风险相关的五个研究空缺。
• 基于AI的信用风险检测模型使用SVM、RF、DT和LR，RF达到约89%的准确率。
• RF的精确度约为88%、召回率约为89%、特异性约为89%、F1约为88%，AUC为0.77。
• 该模型被提议特别适用于定制化信用风险缓解策略。
• 该研究强调了AI在提升个性化的同时管理风险与合规方面的作用。