Publication of the journal

The section is currently being updated

The nowadays supervised machine learning algorithms use the feature description to classify objects. Such a description may include a great number of features provided the task demands it. In the work the genetic algorithm based feature selection as a part of the software complex of bibliographic data processing has been described. The analysis of the problem situation within the framework of the subject area, related to the feature description size of the bibliographic data objects, has been carried out. A method of solving the given problem due to the genetic algorithm feature selection has been proposed. The paper includes the general principles of the software model and the implementation details in the Python programming language. The problem of feature description and re-learning in bibliographic data processing has been solved, it has been shown that learning and re-learning accelerates without loss of the classification quality. The developed software for genetic algorithm feature selection can be applied within the framework of the software complex for bibliographic data processing.The following results have been obtained during the computational experiment: the number of features used decreased from 26 to 15, and the quality of classification increased by 3 % due to the elimination of features that contribute to retraining.

124498, Moscow, Zelenograd, Bld. 1, Shokin Square, MIET, editorial office of the Journal "Proceedings of Universities. Electronics", room 7231

+7 (499) 734-62-05
magazine@miee.ru