New Approach to Portfolio Creation Using the Minimum Spanning Tree Theory and Its Robust Evaluation

Purpose: The aim of this paper is to describe another possibility of portfolio creation using the minimum spanning tree method. The research contributes to the existing body of knowledge with using and subsequently developing a new approach based on graph theory, which is suitable for an individual investor who wants to create an investment portfolio. Methodology/Approach: The analyzed data is divided into two (disjoint) sets – a training and a testing set. Portfolio comparisons were carried out during the test period, which always followed immediately after the training period and had a length of one year. For the sake of objectivity of the comparison, all proposed portfolios always consist of ten shares of equal weight. Findings: Based on the results from the analysis, we can see that our proposed method offers (on average) the best appreciation of the invested resources and also the least risky investment in terms of relative variability, what could be considered as very attractive from an individual investor’s point of view. Research Limitation/implication: In our paper, we did not consider any fees related to the purchase and holding of financial instruments in the portfolio. For periods with extreme market returns (sharp increase or decrease), the use of Pearson’s correlation coefficient is not appropriate. Originality/Value of paper: The main practical benefit of the research is that it presents and offers an interesting and practical investment strategy for an individual investor who wants to take an active approach to investment.