In 2016 Marcos López de Prado published a Hierarchial Risk Parity (HRP) model. The model clusters covariance matrix of expected returns and allocates portfolio weights among clusters. Compared to traditional mean-variance models the HRP model is numerically more stable, because it does not need to compute the inverse of covariance matrix.
In my thesis I will expand the HRP model by incorporating investors views on expected returns and by introducing algorithm of constraints, which modifies suggested optimal portfolio with users assets allocations constraints.
Based on comparisons between optimization methods I have made, we shall see that HRP algorithm optimizes portfolio similar as selecting minimum variance portfolio on efficient frontier. For investors, wishing to have a conservative portfolio, is this very welcoming, because the HRP algorithm can optimize the portfolio even on singular covariance matrix (which a lot of optimization methods can not do). Besides achieving similar results the portfolio optimized with HRP algorithm was also more disperse.
Based on made comparisons it turns out the modified HRP algorithm is also competitive optimization method. The modified algorithm successfully allocates portfolio weights based on expected returns, which allows for incorporating investors