Recommender Systems: A Comprehensive Review from Traditional Methods to Deep Learning-Based Approaches and Large Language Models

Recommender Systems: A Comprehensive Review from Traditional Methods to Deep Learning-Based Approaches and Large Language Models

yaldamesgar1

1) M.Sc. in Computer Engineering (Software Engineering), Islamic Azad University, Ayatollah Amoli Branch

IOI Code : AAHD-ABEAFK
Publication : National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
    Abstract Recommender systems, as one of the most important information filtering tools in the digital age, play an unparalleled role in content personalization and improving user experience. This paper provides a comprehensive review of recommender systems and traces their evolution from traditional methods such as collaborative filtering, content-based filtering, and hybrid systems to advanced approaches based on deep learning, graph neural networks, reinforcement learning, and large language models. In this paper, the mathematical foundations of key methods including matrix factorization, cosine similarity, factorization machines, and convolutional neural networks are precisely described. Furthermore, fundamental challenges such as the cold-start problem, data sparsity, and filter bubbles are examined, and emerging trends including large language model-based recommender systems and agent-oriented systems are analyzed. By presenting a structured classification and a comprehensive survey of the literature, this paper can serve as a reliable resource for researchers and students in the fields of artificial intelligence and information retrieval.

Keywords : recommender systems collaborative filtering matrix factorization deep learning large language models personalization