Evolutive Algorithms with Restrictions Based in Decision Trees Models for the Enhancement of Satisfaction in Time Use

Abstract

A hybrid intelligent system is presented in which the parameters of trained decision trees-based models can be used to define a search space that anoth-er intelligent algorithm can utilize to optimize an objective function. This sys-tem is of value in social sciences research and industry applications dealing with datasets with categorical attributes and non-linear and complex interac-tions, interested in the question of what changes in the realities represented by the datapoints would bring them to a more desirable class or regression value according to a decision maker or policy. The approach has the ad-vantage of high interpretability compared to other black box type intelligent algorithms. A case study is presented in which a dataset with 30 attributes is used to explore the less costly changes in time-use assignations to bring about satisfaction in time use in academic activities.