Knowledge Graph Enhanced Multi-Task Learning for Sequential Recommendation

In the evolving landscape of sequential recommendation systems, this paper propels the frontier forward with the introduction of the knowledge graph enhanced multi-task learning (KGML) model. At its core, KGML harnesses the capability of big data analytics, enabling a nuanced understanding of both the immediate and enduring interests of users. This is achieved through the integration of item knowledge graphs and multitask learning, which are meticulously enriched with big data insights, thereby ensuring a comprehensive representation of item attributes and interconnections. Such a method not only elevates the model's precision in tailoring recommendations for less popular items with limited data but also effectively counters the "Matthew Effect", where visibility becomes disproportionately skewed towards already popular items. Through rigorous validation across three public datasets, the KGML model demonstrates that the proposed approach significantly enhances the accuracy of sequential recommendations.