In the era of Web 3.0, there is a pressing need for a knowledge-centric model to recommend recipes, as the current structure of the worldwide Web lacks specialized recommendation frameworks for recipe recommendations, even though standard search engines offer some level of recipe recommendation. This paper aims to address this gap by proposing a dedicated recommendation channel. The model utilizes user historical activity on food apps as one of the criteria for personalization, leveraging the SimRank algorithm and Grey Wolf optimization to generate a comprehensive termset. Additionally, the model incorporates semantic similarity-based reasoning, employing techniques such as SimRank, Morisita overlap index, and normalized information distance. The Grey Wolf optimization framework further refines the solution set. The topic network comprises recipe ebooks, food blogs, and crawl labels from food blogs, with the inclusion of scientific ebooks to enrich the knowledge base. A restriction classifier is employed to control the features of the data center, ensuring computational feasibility and reducing complexity. Furthermore, a lightweight decision tree classifier demonstrates strong classification performance, characterized by high precision, a lower false discovery rate (FDR), and notable expressive percentage values. Extensive evaluations reveal that the proposed framework outperforms baseline models, establishing it as the leading model for recipe recommendations.