Enhancing Hotel Management Through Predictive AI Models for Customer Lifetime Value (CLV)

Abstract

This paper introduces a hybrid artificial intelligence (AI) framework for predicting Customer Lifetime Value (CLV) in hotel management. CLV represents the long-term financial contribution of guests and provides information for resource allocation, customer retention, and profitability. Traditional models rely on structured reservation records, often overlooking emotional insights in online reviews. To address this concern, the study combines structured booking attributes with unstructured guest reviews, using RoBERTa embeddings for textual data and XGBoost for numerical features. The proposed multimodal model achieves 89% accuracy, 87% precision, and 86% recall, outperforming single-source approaches utilizing bookings (63%) or guest reviews only (72%). SHAP-based interpretability reveals that review topics, including cleanliness, staff professionalism, and service quality, directly influence CLV, alongside structured features such as repeat bookings and special requests. The findings highlight the potential of predictive AI to enhance hotel management by identifying valuable customers early, supporting personalized services, and optimizing strategic decision-making.