INFORMATION-EXTREME MACHINE LEARNING OF KNOWLEDGE CONTROL SYSTEM
Abstract
The algorithm of machine learning of an automated subsystem of control of students’ knowledge according to the test tasks of the computerized education system is considered. In this case, machine learning is carried out within the framework of information-extreme intellectual technology of data analysis, which is based on maximizing the information capacity of the system in the process of its improvement. The results of students' answers to test tasks were considered as signs of recognition, which were evaluated on a scale from 0 to 100. The algorithm of information-extreme machine learning with parallel-sequential optimization of the system of control tolerances on recognition signs was suggested. The lower control tolerance for recognition attributes, with a fixed upper control tolerance, was considered as an optimized machine learning parameter. In this case, quasi-optimal control tolerances on the signs of recognition, obtained in the process of parallel optimization, were used as the starting point for the implementation of a machine learning algorithm with sequential optimization. Kullback modified information measure, which is a function of the exact characteristics of classification decisions, was considered as an optimization criterion of machine learning characteristics. Since the specificity of knowledge control is that the class alphabet is structured, so the enclosed structure of container classes of recognition, which characterize the corresponding levels of knowledge, was considered. In this case, the enclosed structure was characterized by the general center of scattering of vector-realization classes of recognition. This structure, in contrast to polymodial containers of recognition classes, has allowed increasing of the efficiency of machine learning and the validity of decisive rules. The verification of the workability of the suggested algorithm for machine learning was carried out on the basis of a representative input matrix, which was formed on the basis of the student test results of the discipline.
Keywords
