Long Term Optimal Portfolio Selection Problem in Different Models

Date of Award

Fall 9-15-2016

Degree Type


Degree Name

Master of Science (MS)



First Advisor

Dr. Roman Makarov


In this project, we mainly focus on how to set up a complete methodology for finding the best investment portfolio which is meeting investor's risk preferences. In order to consider different aspects of their risk preferences, we use a combination of several risk and performance metrics in the ranking function. After applying this ranking system to select most suitable funds from a large pool, we use different models to optimize the portfolio. Past 5-10 years data set has been extracted from Bloomberg. We find out that the asymmetric multivariate t-distribution can dominate the multivariate normal distribution and the symmetric t-distribution in capturing the market behavior. This has been demonstrated by calculating the likelihood function for each model. We also realize that using a linear combination of several risk or performance metrics can help us to select the most suitable portfolios lying on the efficient frontier that are meeting the unique risk preferences of different investors.

This document is currently not available here.