Searching for the most significant rules: an evolutionary approach for subgroup discovery

In this paper, a new genetic algorithm (GAR-SD$$^{+})$$+) for subgroup discovery tasks is described. The main feature of this new method is that it can work with both discrete and continuous attributes without previous discretization. The ranges of numeric attributes are obtained in the rules induction process itself. In this way, we ensure that these intervals are the most suitable for maximizing the quality measures. An experimental study was carried out to verify the performance of the method. GAR-SD$$^{+}$$+ was compared with other subgroup discovery methods by evaluating certain measures (such as number of rules, number of attributes, significance, unusualness, support and confidence). For subgroup discovery tasks, GAR-SD$$^{+}$$+ obtained good results compared with existing algorithms.