TopRecipes: A Topic Modelling Scheme Driven Framework for Strategic Recipe Recommendation Using Differential Semantics

Abstract

An advanced Web 3.0 framework, exemplified by the Semantic Web, is essential for providing recipes that reflect diverse cultural and geographical culinary traditions within an integrated digital environment. Topic model-based framework, named TopRecipes, encompasses Latent Dirichlet Allocation (LDA) and standard knowledge repositories to generate the informative terms and further depends on the existing knowledge stack to populate the informative terms utilizing TF-IDF. Furthermore, a Semantic network is derived from the terms and strategic standard models like the Normalized Information Distance (NID), and VL and Ahmed index have been incorporated for reasoning based on semantic similarity, differential thresholds, and differential step deviance measures. A Logistic Regression (LR) classifier, a feature-controlled classifier, classifies the dataset, Renyi entropy facilitates feature selection, and dolphin’s echolocation transforms the initial feasible set into an optimal solution set through metaheuristics optimistic computation. The average precision of the proposed TopRecipes was 95.58 with a recall of 97.84, an accuracy of 96.71, and an F-measure of 96.7, where the lowest FDR was achieved, which is 0.05.