Multi-class Object Classification Model Based on Error-Correcting Output Codes

Abstract

The subject of this article is devoted to the problem of object classification. It's introduces a software model for the multi-class classification of image objects based on machine learning algorithms. The results of two modes model operation: OneVsRest and Error Correcting Output Codes are presented. Satellite images of the earth's surface were used as datasets for classification. A comparative analysis of precision, recall, accuracy, hamming loss and classification duration has been carried out. Optimal code rates for Error Correction Output Codes are experimentally determined to achieve increase classification accuracy.