In recent years, hierarchical clustering has been proposed as a method of portfolio construction. It is a method based in machine learning and graph theory. There are many different configurations in which portfolios can be constructed through the use of hierarchical clustering, not all of which have been explored yet. In particular, portfolios constructed through hierarchical clustering have been compared with riskbased portfolio construction techniques in terms of their performance, however, the two approaches have not been combined yet. In this thesis, we go through the steps of portfolio construction through hierarchical clustering and explore how a combination with risk-based portfolio construction techniques could look like. To achieve this, we discuss risk-based portfolio construction techniques, correlation matrix estimation and clustering algorithms separately. Through theoretical results and empirical study, we choose a method for the correlation matrix estimation and two clustering algorithms. Then, we bring these methods together with different risk-based portfolio construction techniques and investigate the performance.     «

In recent years, hierarchical clustering has been proposed as a method of portfolio construction. It is a method based in machine learning and graph theory. There are many different configurations in which portfolios can be constructed through the use of hierarchical clustering, not all of which have been explored yet. In particular, portfolios constructed through hierarchical clustering have been compared with riskbased portfolio construction techniques in terms of their performance, however,...     »