Prediction of Individual Eating Habits Using Short-Term Food Recording

In order to maintain and improve health, it is very important to grasp one's eating habits in his/her everyday life. Ideally, individual long-term food recording is preferred to make health suggestion about eating. However, it is not easy for an individual to continuously record his/her dishes in the long term. It would be beneficial to predict one's eating habits only by his/her short-term food recording. In this paper, we propose a method to predict individual eating habits only by short-term recording. We make use of personal food recording data and methodologies of recommendation such as collaborative filtering and evaluating the accuracy of prediction. In experiments using various terms of recording, we are able to achieve an accuracy of more than 70% for details about food dishes, and more than 90% in a food category using seven days of food recording.