Modeling Artificial Emotional Agent Based on Combination of Classification and Fuzzy Logic Methods

As life is being more digitalized every day, scientists try to design intelligent systems, which do not interact like machines or in pure logical ways without flexibility, soul and unnatural. In fact, artificial intelligent systems that act in novel ways like human-human interaction for social interactions between computers or robots with humans, creatures, and environments. In this study, we propose a model to simulate an artificial emotional agent by using classification and fuzzy logic methods. Since it is difficult to determine just a dominant feeling for a sentence, six basic emotions are defined to figure out the dominant feeling by applying various multi-classification algorithms to find out the optimized kernel and classifier. Then the classified sentence sent to a fuzzy logic system to expand the emotions and achieve a variety of fuzzy emotions based on the sixteen distinct emotions of the Geneva emotional wheel. Besides, the lingual variable, valence and control values achieved as the system outputs that make the experiment a successful step to simulate natural human emotions. In this study, we succeed to simulate humanoid emotions based on the input sentences and psychological-emotional models with 80% accuracy by multi-classifier algorithms and Mamdani fuzzy expert system.