RECOMMENDER SYSTEMS IN SOCIAL NETWORKS USING CROWDSOURCING

Abstract

The rapid growth of content in social networks has made it hard for users to find relevant information and has raised the need for effective recommender systems. Collaborative filtering is one of the most widely used recommendation approaches, yet it faces the challenges of data sparsity and cold start. The aim of this study was to design and evaluate a hybrid recommender system that strengthens collaborative filtering with crowdsourced data. The study followed an applied and developmental design, and the data were gathered from an active social-network group in the health-news domain. The implicit data were drawn from the users' behavioral traces, and the explicit data were obtained from a researcher-made questionnaire across four dimensions, namely satisfaction, relevance, transparency, and trust. The model performance was measured with precision, recall, the F1 score, coverage, diversity, and prediction error. The findings showed that crowdsourcing increases data richness and density and raises the density of the interaction matrix from 6.3 percent to 28.9 percent. The structured integration of explicit and implicit data also lifted the precision of the optimized version to 0.45 at the top ten recommendations and its recall to 0.44, while lowering the prediction error. These improvements enhanced recommendation quality for low-activity users and eased the cold-start problem. Overall, integrating crowdsourcing with collaborative filtering provides a human-centered and effective framework for improving recommendation in social networks.