Predicting wine quality from terrain characteristics with regression trees

Abstract

A former cartographic study on terrain characteristics of the German Rhinegau is reviewed. An attempt is made to predict relative quality of the Riesling from local site factors usings Classifcation And Regression Trees (= CART). Valid results suppose quantity to be ruled out by quality, ignoring any price ratio of vine cultivation. The study demonstrates that CART is a valuable statistical tool without restrictions by data types.

Keywords

• regression classification
• agriculture
• wine quality
• terrain characteristic
• statistics
• vineyard ecology