Knowledge Testing System Base on Machine Learning and Fuzzy Systems

The paper discusses a method for constructing systems for assessing the quality of knowledge, in particular testing systems, based on neural networks and fuzzy inference systems in order to improve the accuracy and objectivity of the assessment. A corresponding three-layer architecture of the estimation model has been developed. An evaluation system is proposed based on two-criteria evaluation of (correct and incorrect) answers with further processing of the obtained data by fuzzy inference methods to obtain the final result. A two-stage approach was developed to improve the quality of knowledge control, the use of a neural network allows you to find relationships between a set of answers and the found degrees of correct and incorrect answers. The system allows you to comprehensively consider the test, linking knowledge on the topics studied, thereby assessing the overall level of the student.