Loading...

Optimal Distance Calculation Method for Portfolio Optimization using
Nested Cluster Optimization

Rafatnezhad, Ramtin | 2023

79 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56186 (44)
  4. University: Sharif University of Technology
  5. Department: Management and Economics
  6. Advisor(s): Arian, Hamid Reza; Zamani, Shiva
  7. Abstract:
  8. In the basic model of this thesis, which is called nested cluster optimization, only one distance function is used for clustering to form clusters with similar characteristics, while depending on whether the optimization model is long-only or long-short, different functions can be used. The aim of this thesis is to find the optimal distance function between assets in the simple nested cluster optimization so that during three different and separate strategies, based on three criteria of the lowest risk, the highest Sharpe ratio, and the highest return, the optimal distance function of assets is selected and clustering and finally weighting the portfolio to be done. The optimal distance function between assets is selected monthly in the training data using a rolling sample so that we finally have a monthly series of asset weights and monthly performance of financial portfolios. In the following, the performance of the financial portfolios in these three developed methods will be compared with the performance of the common methods of choosing the optimal portfolio that is introduced, in the criteria of return, risk, Sharpe ratio, certainty-equivalent (CEQ) return, transaction costs, and turnover, so that a correct assessment of the performance of the developed method can be made. Algorithms run on two data sets in stocks and cryptocurrencies. In the long-only mode, the performance of the developed nested cluster optimization method has been better in the stock data according to the risk and Sharpe ratio, while in the cryptocurrency data, the result has been significant in the Sharpe ratio criterion. But in the long-short, the performance of the developed nested cluster optimization method in the stock data was better than the simple method of this algorithm; But in the cryptocurrency data, we observed a much better performance in the Sharpe ratio criterion despite the lower turnover
  9. Keywords:
  10. Portfolio Optimization ; Clustering ; Machine Learning ; Optimization ; Nested Clustering Optimization ; Asset Allocation

 Digital Object List