Purpose 

Because many health conditions can be attributed to a bad diet, there is a clear need for a mechanism to help people make better nutritional choices. Food and nutrition need to be incorporated into healthcare. For people with heart disease, high blood pressure, diabetes, or food allergies, a recipe recommender system for improved health could analyze the nutritional requirements critical to managing their condition and suggest recipes accordingly. This recommender system could not only provide the solution to meet an individual’s health needs but also incorporate an individual’s food preferences. Personalized systems that align with patient’s tastes and needs could make following dietary recommendations easier and improve overall health. Doctors, dieticians, hospitals, and home cooks could use such a solution. 

Objectives 

The first objective of this project was to create healthy recipe recommender systems using different techniques. These include knowledge-based, content-based, collaborative filtering, machine-learning algorithms using the Python surprise library, and hybrid methods combining several techniques. The second objective was to show the effectiveness of each type of recommender system with examples using the Food.com dataset and explain the strengths and weaknesses of each type – effectiveness was based on the relevance of the recipes suggested and the calculated RMSE. The third objective was incorporating sentiment analysis of user reviews to determine preferences based on emotional recipe responses. The resulting sentiment scores were used to predict the recipe ratings of users and combined with other methods to assess the best-suggested recipes. Finally, the last objective was to compare the actual recipe ratings with the sentiment scores by calculating precision, recall, and accuracy and demonstrating the potential benefits of sentiment analysis.