Company classification by Non-negative Tensor Factorization

Abstract

It is important for a company's business strategy to understand how customers contact the company and what kind of brand image they have. With the development of information technology in recent years, it has become possible to connect with anyone, anytime, anywhere, and companies are increasingly diversifying their customer contact points and collaborating with different industries. However, the complexity of brand images formed from a wide variety of business strategies has made it difficult to extract the characteristics of each company. In this study, we apply a Non-negative Tensor Factorization to corporate brand survey data to extract characteristic patterns of corporate brand image and customer contact, and classify companies. Furthermore, to identify similar companies, we grasp a position of each company by the dimensionality reduction of extracted feature matrix.