THE DEVELOPMENT OF AGENT-ORIENTED INTELLIGENT TRAINING SYSTEM BASED ON NEURO-FUZZY TAKAGI-SUGENO-KANG SYSTEM

One of the main requirements to the distributed training systems is the minimal interaction between its components located in various computers; the provision of variations of computer technologies studying. In this respect, the authors proved that the development of intelligent training systems should be based on a syncretical solution of the problems in the sphere of computer technologies and didactics.

This work proved the applicability of agent-oriented approach while developing new generation intelligent training systems that allows to implement specific algorithms of interaction of intelligent training systems separate components in case of the system unspecified algorithm in the whole.

Basing on the analysis of mathematical methods used while developing training systems the authors have chosen the method based on the integration of several mathematical methods, such as neuro-fuzzy Takagi-Sugeno-Kang system. It carries out training using the genetic algorithm.

The authors introduced the methodology of development of functioning algorithms of intelligent agents which includes the following operating actions: the definition of input and output variables; the selection of model for the rule base derivation and formation; the selection of fuzzification function and the definition of training set parameters; the definition of belonging of training set elements to fuzzy rules; the fuzzification layer parameters settings; the conclusion layer parameters settings. The authors developed the functioning algorithm of a subsystem of the intelligent behavior of a correcting agent of individual educational course. The developed algorithm realizes the variations of free choice degree of the learner’s individual educational course depending on the current monitoring results, the solved problems difficulty level and the quantity of prompts given while solving problems. The authors presented the fragments of settings and testing of the developed algorithm.

Basing on the developed algorithm of functioning of subsystem of intelligent agent’s behaviour the authors created the intelligent training system on the discipline “Fundamental management” which efficiency has been proved by the students of Orenburg state university.