Title: Exploring the Synergy of Topic Modeling and Prefixspan Algorithm in Developing a Hybrid Recommender System for Social Media Platforms
Journal of Artificial Intelligence and Data Science Techniques
© 2024 by jaidst - PROVINCE Publications
ISSN: 3029-2794
Volume 01, Issue 03
Year of Publication : 2024
Page: [27 - 44]
Ramesh Krishnamurthy and Priya Mehta
Department of Machine Learning, Indian Institute of Technology, Bombay, India
Department of Software Engineering, Banaras Hindu University, India
Content creation by users on social media platforms has increased exponentially. Without a recommender system, creating relevant and personalised material is tricky. The ever-changing material and user preferences make it difficult for traditional recommendation systems to keep up. To address these issues, this work proposes TopiXscan, a novel hybrid recommendation system that combines topic modelling with the Prefixspan technique. Latent Dirichlet's Allocation (LDA) and other topic modelling approaches are used by the TopiXscan model to extract latent topics from user-generated content. As a result, user preferences and material quality may be explained semantically. Prefixscan, an ordered pattern extraction tool, may be able to capture the brief changes in user behaviour and analyze their interactions with common sequence patterns, according to the study. To make the most of both fields, the TopiXscan model built a hybrid engine for recommendations that used content-based and collaborative filtering techniques. If the application wants to know what the user values most, it may model more than just their hobbies and interests to provide personalized content suggestions. However, Prefixscan will keep tabs on what users do and then use that data to tailor content recommendations to their changing tastes. Real-world social media datasets will be analyzed to test how well the proposed hybrid recommendation system works. The findings demonstrate that the latter outperforms conventional recommendation systems when it comes to of variety, uncertainty, and accuracy. Furthermore, the study showcased the potential synergy between topic modelling and a sequential information mining technique to improve quality in high-information, dynamic environments.
Prefixscan algorithm, TopiXscan model, Topic modelling, Sequential pattern-mining. Hybrid recommender system, Latent Dirichlet Algorithm, Social media platforms.