Typical recommender systems try to mimic the past behaviors of users to make future recommendations. For example, in food recommendations, they tend to recommend the foods the user prefers. While the recommended foods may be easily accepted by the user, it cannot improve the user’s dietary habits for a specific goal such as weight control. In this presentation, we describe an approach to food recommendation trying to take into account the goal of the user (e.g. weight loss), the health information (BMI) of the user and the nutrition information of foods (calories). Instead of applying dietary guidelines as constraints, we build recommendation models from the successful behaviors of comparable users: the weight loss model is trained using the historical food consumption data of similar users who successfully lost weight. We tested the approach. Experiments on real data collected from a popular weight management app show that this recommendation approach can predict more appropriate foods than a traditional recommendation approach.

Recording: https://drive.google.com/file/d/1EHmI3hFgwP_6PzYG2KqERgxssRwHXY4H/view?usp=sharing