Clustering companies based on the time series of stock returns

Abstract

In this paper, we propose a method of clustering companies using stock price time-series data. Some companies have multiple businesses and are not limited to a single industry. However, in the conventional industry classification, they are classified into any one industry, and the industry classification differs from the actual situation. In addition, some different business categories are gathered into one group. We have proposed a method of clustering index-type mutual funds using fund returns time-series data. This paper extends this method to companies and stock returns. We apply the proposed method to actual stock price data and confirm its effectiveness.