Portfolio Creation Using Graph Characteristics and Testing Its Performance

ABSTRACT

The aim of this paper is to propose a method for selecting underlying assets for the investment portfolio so that we can achieve a maximum expected return with minimal risk. For this purpose, we describe the portfolio creation process using the minimum spanning tree method, a graph theory tool. Using this theory, we select individual stocks from the Dow Jones Industrial Average index which we include in the portfolio and then compare the performance of such a portfolio with three alternative investments. The first alternative investment is a purchase of the entire Dow Jones Industrial Average index, the second alternative investment is to create a portfolio according to the Markowitz portfolio theory, and the third alternative investment is to create a portfolio through a random selection of indexforming stocks. The comparison was made mainly with regard to the average annual return and volatility of portfolios. In order to ensure the objectivity of the results, we divided the data into two disjoint sets. One of them was a training set, on which we built the portfolio, and the other one was a testing set used for performance testing. In the article we present that our portfolio can achieve results at least as good as those of the given alternative investments. In particular, this method also enables individual investors to determine how many underlying assets should be included in the portfolio.