During the last decades, researchers in the field of Term Rewriting System (TRS) have devoted a lot of effort in order to develop techniques and methods able to demonstrate the termination property of a TRS. As a consequence, some of the proposed techniques have been implemented and several termination tools have been developed in order to automatize the termination proofs. From 2004, the annual Termination Competition is the foro in which research groups compare their tools trying to provide termination proofs of as many TRS as possible. This event generates a large amount of information (results obtained by the different tools, time spent on each proof, …) that is recorded in databases. In this paper, we propose an alternative approach to study the termination of TRS: to use data mining techniques that, based on the historical information collected in the competition, generate models to explore the termination of a TRS. The goal of our study is not to develop a termination tool but to show, for the first time, what machine learning techniques can offer to the analysis of TRS termination.