Illuminating the black box: Explainable AI for enhanced customer behavior prediction and trust
Year: 2025
Type: article
Abstract: Artificial intelligence is increasingly used in business, particularly for predicting customer actions, which has improved forecasting accuracy. However, powerful machine learning models are often complex "black boxes," hiding the reasons behind predictions. This lack of visibility limits actionable insights, complicates meeting regulations like GDPR's 'right to explanation,' and makes building trust with customers and others difficult. Explainable Artificial Intelligence (XAI) addresses this by making AI models more understandable. This paper looks at applying XAI methods to customer behavior prediction (CBP) models. We evaluated model-agnostic XAI techniques, including SHAP and Permutation Feature Importance, on high-performing Gradient Boosting and Deep Learning models. These models were trained on real customer transaction and interaction data. Our analysis focused on predictive performance and the quality of explanations provided by XAI. We found that while complex models often predict better, XAI successfully revealed the key features and interactions influencing their predictions. This yielded insights into the reasons for customer actions. Our work shows how XAItransforms black box predictions into clear, usable intelligence. This helps businesses improve personalization, refine marketing, manage risk, and build customer relationships through openness and confidence. This research contributes to applying XAI, specially for understanding and predicting customer actions.
Authors: Raiyan Haider, Md Farhan Abrar Ibne Bari, Osru, Nishat Afia, Tanjim Karim
Institutions: University of Lancashire, Daffodil International University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4411111802