Construction And Evaluation Of Recipe Recommendation System Considering User Taste Preferences And Nutritional Balance
Currently, there are many cooking and recipe sites. However, they often do not reﬂect a user’s taste preferences nor include detailed nutritional information. In this study, we construct a novel system for recommending food recipes that reﬂect user’s taste preferences and provide a nutritionally balanced set of ingredients, focusing on user’s most and least favorite ingredients contained in the recipe and the nutrients of each ingredient. To recommend an appropriate food recipe, the method ranks food recipes by combining the personal preference score and the nutrition score for each recipe. For this, we ﬁrst calculate the personal preference score for each recipe based on the user’s most favorite and least favorite values for ingredients using his or her browsing history of cooking and recipe sites. We then calculate the nutrient score for each recipe based on the cosine similarity between the feature vectors of the nutritional value of each ingredient and the reference values for dietary intakes. In this way, the system ranks and recommends recipes based on the recipe scores by combining the personal preference score and the nutrient score for each recipe. Finally, we evaluated our proposed recipe recommend system through a user study.
Index Terms - Nutritional balance, Recipe recommendation, System construction and evaluation, User taste preference.