Generation of Consistent Sets of Multi-Label Classification Rules with a Multi-Objective Evolutionary Algorithm

Multi-label classification consists in classifying an instance into two or\nmore classes simultaneously. It is a very challenging task present in many\nreal-world applications, such as classification of biology, image, video,\naudio, and text. Recently, the interest in interpretable classification models\nhas grown, partially as a consequence of regulations such as the General Data\nProtection Regulation. In this context, we propose a multi-objective\nevolutionary algorithm that generates multiple rule-based multi-label\nclassification models, allowing users to choose among models that offer\ndifferent compromises between predictive power and interpretability. An\nimportant contribution of this work is that different from most algorithms,\nwhich usually generate models based on lists (ordered collections) of rules,\nour algorithm generates models based on sets (unordered collections) of rules,\nincreasing interpretability. Also, by employing a conflict avoidance algorithm\nduring the rule-creation, every rule within a given model is guaranteed to be\nconsistent with every other rule in the same model. Thus, no conflict\nresolution strategy is required, evolving simpler models. We conducted\nexperiments on synthetic and real-world datasets and compared our results with\nstate-of-the-art algorithms in terms of predictive performance (F-Score) and\ninterpretability (model size), and demonstrate that our best models had\ncomparable F-Score and smaller model sizes.\n