Multi-Modal Graph-Based Recommendation System: Integrating Heterogeneous Modalities for Enhanced Predictions

In the realm of modern digital applications, recommender systems have emerged as vital tools, assisting users in navigating a plethora of choices. This paper addresses the challenge of enhancing recommendation accuracy by harnessing the capabilities of Graph Neural Networks (GNNs) and multi-modal data integration. GNNs are identified as effective tools for enriching node feature representations by propagating local information. However, their efficacy wanes when dealing with data-scarce nodes. In response, the study introduces a novel approach: a multimodal graph-based recommendation system. This approach synergizes GNNs with diverse modal features, the approach excels in predicting connections within heterogeneous graphs and assessing similarity across nodes. Through the fusion of multiple modalities, including text and images, via advanced feature engineering techniques, the accuracy of node embeddings is enhanced, leading to substantial improvements in recom-mendation precision. The evaluation, conducted on the Neo4j MovieLens dataset, attests to the system's efficacy in harnessing varied data sources. The concept of a unified node representation further bolsters the system's proficiency in capturing latent semantic relationships between nodes. This integration is instru-mental in enabling content-based recommendations, showcasing the model's competence in encapsulating user-item interactions and rendering accurate movie rating predictions. Ultimately, this research sheds light on the synergy between graph-based machine learning and recommendation systems, accentuating the pivotal role of data modalities in enhancing accuracy and personalization. The proposed approach holds the potential for fostering robust recommendation systems across diverse domains.