Open Access

ARTICLE

Cognitive-Based Enhanced Accuracy and Relevance in Cross-Domain Recommendations

Luong Vuong Nguyen^1,*, Hoang Tran², Thuy-Trang Pham³

1 Faculty of Artificial Intelligence, FPT University, Danang, Vietnam
2 Department of Software Engineering, FPT University, Danang, Vietnam
3 Department of Business, FPT University, Danang, Vietnam

* Corresponding Author: Luong Vuong Nguyen. Email:

(This article belongs to the Special Issue: Next-Generation Recommender Systems: Multimodality, Generative Models, and Trustworthy Personalization)

Computers, Materials & Continua 2026, 88(3), 92 https://doi.org/10.32604/cmc.2026.082406

Received 15 March 2026; Accepted 09 June 2026; Issue published 23 July 2026

Abstract

In the era of information overload, cross-domain recommendations offer a promising solution by leveraging user preferences across domains to improve recommendation accuracy and relevance. This study proposes a novel approach to cross-domain recommendations based on cognitive similarity derived from user-based features. We construct comprehensive user profiles across multiple domains by defining cognitive similarity based on user interaction data, including ratings, reviews, and genre preferences. We employ advanced feature extraction techniques, including TF-IDF for textual data and matrix factorization for latent factors, to quantify similarities in user preferences across domains. These cognitive similarity measures are then used to map user profiles into a common latent space, facilitating the generation of personalized cross-domain recommendations. To visualize the effectiveness of our approach, we use methods such as Multidimensional Scaling (MDS) and heatmaps to depict the cognitive similarity between users across different domains. Additionally, network graphs illustrate the intricate relationships and similarities across user profiles, offering intuitive insights into the recommendation process. The results demonstrate that our cognitive similarity-based approach significantly improves the relevance and diversity of cross-domain recommendations, providing a robust framework for future research and practical applications in personalized recommendation systems.

---

Keywords

---