Ensembles of Decision Trees for Recommending Touristic Items

This article analyzes the performance of ensembles of decision trees when applied to the task of recommending tourist items. The motivation comes from the fact that there is an increasing need to explain why a website is recommending some items and not others. The combination of decision trees and ensemble learning is therefore a good way to provide both interpretability and accuracy performance. The results demonstrate the superior performance of ensembles when compared to single decision tree approaches. However, basic colaborative filtering methods seem to perform better than ensembles in our dataset. The study suggests that the number of available features is a key aspect in order to get the true potential of this type of ensembles.