Analyzing the attractiveness of brand names using machine learning

Abstract

The article is devoted to analyzing the problem of identifying hidden regularities in the mechanism of society's reaction to brands using machine learning methods to assess the attractiveness of their names. To solve this problem, the authors conduct a study involving the analysis of a large amount of data containing various brand names divided into spheres and information about the degree of their popularity among consumers by applying mathematical methods and machine learning algorithms. In this study, experiments using static Word2Vec predictive models with three different corpora were conducted to validate the effectiveness of the proposed approach. The results of the study demonstrate that the approach developed by the authors allows finding names that meet the requirements of the business domain based on consumers' evaluations.