Improving the Recommendations of Meals in the PROMISS Application

The PROMISS application is specifically built to let older adults keep track of their diet and protein intake. To improve the user-experience of this application, we study how machine learning algorithms can be used to recommend meals and products based on historical data. An intelligent workflow is designed which combines five different algorithms that recommend suitable meals and products. These algorithms are trained and tested using data from a previous user study with the PROMISS application. The change in user-experience is measured by the numbers of clicks needed to enter a meal in the application. Two different variants of the new application, namely, one using only the two new recommended meals and the other using both the two new recommended meals plus the old recommended meal, are compared with the old application. It was found that both new applications reduce the number of clicks and thus increase the user-experience of the application.