Comparison of two strategies, CTC and CMM, to combine m classifiers in a single comprehensible one

Accurate prediction is probably the most pursued objective when solving real problems with machine learning, but there are situations where, added to the prediction, it is important to obtain a comprehensible output. The aim of this work is to compare the behaviour of two strategies to combine the knowledge of m classifiers in a single one in order to maintain the explaining capacity of the final classier: Consolidated Tree’s Construction (CTC) algorithm and Combined Multiple Models (CMM) algorithm. The comparison is done from three points of view: accuracy, complexity and stability in explanation. Experimental results show that the use of CTC would be more recommendable than the use of CMM because, even if from the accuracy point of view the behaviour of CTC and CMM is similar, CT trees will give a more comprehensible (59.2% simpler) and steadier explanation (structure 50% steadier) than CMM classifiers.