RMIKD: An RDF and Metadata-Driven Scheme for Recommending Web Images for E-Commerce Fashion and Products Using Incremental Knowledge Derivation

In this paper, the RMIKD framework, which is an RDF and metadata-driven framework for Web image recommendation, has been proposed that encompasses latency indexing and structural non-topic model for discovering and harnessing topics into the localized framework. Subsequently, the dataset is classified using the XGboost algorithm, which is a feature control machine learning but a strong framework that prevents deviation of the recommendation from the constituent domain and subsequently the framework generates RDF. The proposed framework generates RDF by using the RDF subject and object in order to end provision, strong label co-occurrence, turn semantics and generate ontologies to provide strong knowledge representation models in order to increase density of auxiliary knowledge into the framework. And most importantly, the framework also generates metadata and classifies it using transformers, which eases the overall handling of metadata and curates it as individual instances or capsules, and most importantly, a strong relevance computation scheme in terms of The Leacock Chodorow similarity, Normalized Compression Distance (NCD) and Sim Rank along with the Normalized Information Distance (NID) is being constituted at different parts or at different stages in the proposed RMIKD framework, which provides a strong ecosystem of computing semantic entity driven, semantic relatedness. The proposed RMIKD yields an overall average precision percentage of 97.07, average accuracy percentages of 97.88 and lowest FDR of 0.03.