Development of the Program for the GRNTI Matching According to the Text, Articles and Keywords

Text classification algorithms of machine learning were observed and researched this final qualifying work. The core purpose of this work is developing application for classification scientific articles by GRNTY code. Learning dataset based on articles from e-library “Cyberleninka”. By this documents landmark attributes were selected name of article, GRNTY code, abstract and text of article. Text of articles transform to matrixes of features by text processing algorithms. Classification models, such as Bayesian classification, logistic regression, random forest classification, SVM, were built by arrays of features. Relations between quality of learning algorithms and characteristics of learning dataset were identified. The best method was added in processing pipe and save in special file. Application identifying GRNTY code was created by using the pipe from file. This application will be used for classification articles automatically. Volume of the final qualifying work is 63 pages (including the appendix – 70 pages), the number of illustrations – 28, number of tables – 5, number of references – 49.