Predicting food demand in food courts by decision tree approaches

Fluctuations and unpredictability in food demand generally cause problems in economic point of view in public food courts. In this study, to overcome this problem and predict actual consumption demand for a specified menu in a selected date, three decision tree methods (CART, CHAID and Microsoft Decision Trees) are utilized. A two year period dataset which is gathered from food courts of Hacettepe University in Turkey is used during the analyses. As a result, prediction accuracies up to 0.83 in R2 are achieved. By this study, it’s shown that decision tree methodology is suitable for food consumption prediction.