Personalized Recipe and Meal Planning System Using Cosine Similarity and KNN – Based Similarity Ranking

Abstract

In the realm of personalized cooking experiences, improving the methods of finding and organizing recipes can enhance the cooking experience for a user. This research presents a Personalized Recipe Recommendation and Meal Planning System that supports recipe recommendations based on ingredients, cuisine, course, and dietary restrictions. Our system implements contemporary machine learning techniques, such as TF-IDF vectorization, cosine similarity, and KNN-based similarity ranking in order to develop intelligent matches from user input and ingredients while providing recommendations from a wide range of recipes. In addition to real-time personalized recommendations, we develop a meal planning module to help users more efficiently manage their weekly meals by providing a dynamically generated meal plan structure of their meals based on the user`s preferences. Our web-based interface is designed using Streamlit to provide a smooth and interactive user experience. Data normalization and feature encoding are used to ensure that similarity measurements are balanced and accurately represent user input and recipes across several types. Our experiments include results to showcase that our system was able to adapt to a user`s preferences, return highly relevant real-life recipe recommendations, while allowing users to better organize their weekly meals - in turn enhancing both the convenience and enjoyment of cooking at home.