Automatic generation and recommendation of recipes based on outlier analysis

Research results on medicine and health show that people nowadays tend to have some common diseases because of abnormal eating habits, irregular lifestyles, fast-food culture, etc. Diabetes and high blood pressure are just two examples. This study is based on an ontology-based dietary management system established by our group earlier. The main contribution of this paper is to propose a method for synthesizing new recipes based on existing ones, and recommending proper recipes based on machine learning. The new recipes are combinations of several existing ones. They are recommended to the user only if necessary nutritions are properly contained in the recipe. Outlier analysis is used to judge if a recipe is good or not. Some primary experiments are conducted to show the usefulness of the proposed method.