Portfolio selection with self-organizing maps: An application in BIST 100

The most important stage of portfolio management is diversification. With a good diversification, the risk of the portfolio can be minimized and the most suitable portfolio can be obtained for the investor. Not only the risks and returns of the stocks but also some financial indicators that affect the risks and returns of stocks (price/earnings ratio, market value/book value, earnings per share, equity/debt ratio etc.) are of great importance in order to make an effective diversification. The purpose of this study is to find out similar and dissimilar stocks and to make effective clustering of the stocks included in BIST-100 index by employing their financial indicators that reflect their specific characteristics as the inputs. With clustering analysis, stocks with similar characteristics are gathered together to construct the effective portfolios. In the first two sections of the study, there are concepts such as portfolio, portfolio management, portfolio return and risk, portfolio management approaches, self-organizing maps, cluster analysis. In the application section of the study, 94 stocks which are traded in the BIST-100 Index between 2014 and 2018, were divided into homogeneous clusters by using the self-organizing maps method by using 11 financial indicators as a variable about stocks. The best-performing stocks are determined according to their return/risk ratios in the cluster and optimum portfolios are constructed to take into consideration of the risk-averse and risk-neutral investors. KEYWORDS: Self-Organizing Maps, Cluster Analysis, Portfolio Management, Portfolio Optimization, BIST-100 Index