Portfolio Optimization with Clustering Methods

The portfolio selection problem and optimizing it is an important issue in investment field. This process aims to create A portfolio that incur the minimum amount of risk to the investor in different levels of return. The existence of statistical uncertainty and noise in information, which is used in the process of optimization, would hamper the performance of optimized portfolios. The purpose of this article is to eliminate the noise of correlation coefficient matrix by using clustering method. In order to do so, two clustering methods, Single Linkage and Average Linkage, are applied, and the optimized portfolio based on the Tehran Stock Exchange information from March 2007 to March 2013 is obtained. The results show that such portfolio is more reliable and provides less risk to the capital owner. For analyzing the robustness of the results in a wide range of different investment horizons and numbers of assets, a Bootstrap analysis with resampling is performed and in most cases a significant improvement was observed.