This repo pertains to the work developed for a portfolio optimization research project. Authors are: Luiz Fernando Medeiros (Maasalo-Medeiros Oy) and Professor Esa Ollila (Aalto University). An initial version of the research details and results may be found in the following link.
This work evaluates the performance of different estimators in the domain of Portfolio Optimization. Three different estimators, namely sample based, Ledoit-Wolf (LWE), and Rotationally Invariant Estimator (RIE) are evaluated on their ability to approximate Covariance matrices, as well as how these Covariance matrix approximation affect the portfolio optimization process. The general conclusion is that LWE outperforms SCE and RIE under most conditions, while RIE outperforms in a few different scenarios.