A Novel Methodological Approach to Structural Integration in Recommender Systems and Improve Interaction Modeling by Graph Neural NetworksA Novel Methodological Approach to Structural Integration in Recommender Systems and Improve Interaction Modeling by Graph Neural Networks

A Novel Methodological Approach to StructuralIntegration in Recommender Systems and ImproveInteraction Modeling by Graph Neural NetworksNowadays, recommender applications are necessary in information filtering,
by suggesting priorities to users. Traditional methods often struggle with
implementation defects. Generally, condition inconsistencies and dynamic
variables are ignored in most existing Recommender Systems (RSs). Such
challenges are due to the lack of the holistic, dynamic and integrated
approach in the development process of RSs. In order to deal with such
challenges, key achievements of this research can be expressed from two
dimension. First, this study addresses such limitations by designing a novel
methodological approach toward holistic, dynamic and extensible
development process for RSs. Afterward, in order to express the applicability
of the proposed approach and prove its validity, a novel recommender
framework based on the Graph Neural Network (GNN) has been developed
and evaluated, through major performance metrics. The proposed
framework evaluated via MovieLens real-world dataset, with important
performance metrics. Results present significant improvements over
compared baseline models. The proposed approach outperforming compared traditional models and enhanced the accuracy and
robustness of recommendations. The ability of the proposed framework to
capture the intricate relationships in the user-item interaction data, can lead
to more accurate and personalized recommendations. Moreover, applying the
proposed approach can lead to structural integrity, operational transparency,
dynamism and expandability of RSs.