Portfolio Optimization Through Wisdom of the Crowd

Abstract

In certain mathematical financial models, stocks are modeled Geometric Brownian motions. In order for these models to have any value, there must be set parameters for expected return and volatility for the stocks in the portfolio. Determining an expected return parameter for a stock is a very difficult task. There are vast amounts of information on the web and elsewhere that make predictions on individual stocks and the markets as a whole, but no one really knows the true expected return of a stock. Stock analysts from big Wall Street firms regularly release one-year price targets on certain stocks, but obviously, their forecasts are not perfect. This thesis will attempt to adjust analysts' price targets on 8 of the largest stocks in the S\&P 500 using machine learning techniques and other adjustments to come up with an expected value parameter for the Geometric Brownian Motion model of a stock. Once the expected value parameter has been calculated, this thesis will test the performance of the growth-optimal portfolio from March 2017 to March 2018.

Furthermore, this thesis will evaluate the performance of growth optimal portfolios using adjustments of the consensus price targets as the expected value parameter on the 15 largest stocks in S\&P. The performance of these portfolios will be tested from January 2015 to March 2018.