Scholarlink

<p>This study investigates the role of Explainable AI (XAI) in improving the accuracy, efficiency, transparency, and trust in decision-making processes in personalized retail-banking recommendation systems. A thorough examination of machine learning models was conducted using a dataset from Kaggle, comprising 15,168 rows and 12 retail banking demographic features. To understand and validate recommendation model, we used methodologies from popular XAI library packages, such as LIME, SHAP, Yellowbrick, ELI5, and Alibi. The collaborative insights provided by PCA techniques resulted in the identification and removal of detrimental features, which significantly improved the model accuracy and precision. The Results in graphical representations and analyses provide guided model refinement strategies. The overall model accuracy demonstrated a notable improvement, increasing from 0.86 to 0.91, highlights the effectiveness of incorporating banking-related features within XAI-based recommendation systems. This positive performance trajectory further validates the superiority of the XAI models over their non-XAI counterparts across all evaluated features. In particular, the SHAP-enhanced Random Forest model, under the ensemble learning category, not only achieved higher accuracy but also excelled in terms of interpretability and bias detection capabilities. These results underscore the potential of XAI in banking applications, suggesting a pivotal shift towards models that offer enhanced precision and improved transparency, thereby fostering greater trust and reliability in personalized customer service recommendations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A mini-approach for retail banking with transparent recommendation system enabled by explainable AI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月21日

查看全文
  • Palash Bairagi,
  • Shrey Arora

摘要

This study investigates the role of Explainable AI (XAI) in improving the accuracy, efficiency, transparency, and trust in decision-making processes in personalized retail-banking recommendation systems. A thorough examination of machine learning models was conducted using a dataset from Kaggle, comprising 15,168 rows and 12 retail banking demographic features. To understand and validate recommendation model, we used methodologies from popular XAI library packages, such as LIME, SHAP, Yellowbrick, ELI5, and Alibi. The collaborative insights provided by PCA techniques resulted in the identification and removal of detrimental features, which significantly improved the model accuracy and precision. The Results in graphical representations and analyses provide guided model refinement strategies. The overall model accuracy demonstrated a notable improvement, increasing from 0.86 to 0.91, highlights the effectiveness of incorporating banking-related features within XAI-based recommendation systems. This positive performance trajectory further validates the superiority of the XAI models over their non-XAI counterparts across all evaluated features. In particular, the SHAP-enhanced Random Forest model, under the ensemble learning category, not only achieved higher accuracy but also excelled in terms of interpretability and bias detection capabilities. These results underscore the potential of XAI in banking applications, suggesting a pivotal shift towards models that offer enhanced precision and improved transparency, thereby fostering greater trust and reliability in personalized customer service recommendations.