Portfolio Management with Heuristic Optimization

Abstract

Portfolio Management is the art of choosing and managing a group of financial securities such as bonds or equity instruments, derivative instruments like futures and forward contracts etc. There many key elements to portfolio management, such as asset allocation, diversification of the portfolio, and rebalancing the portfolio. In this project, my objective is to develop strategies to build optimal portfolios which aim to outperform index funds such as the S&P500 Index by researching and developing different heuristic methods for the key elements of portfolio management. One of the strategies use various fundamental factor information such as return on an asset, asset turnover, net margin etc., to build alpha factors. The alpha factors are used as input variables in a machine learning clustering model which is used to diversify and build portfolios from the stock universe. Another strategy uses Piotroski F-scores which is calculated using financial metrics which can be grouped into three criteria's, leverage, profitability and liquidity to filter out companies with bad fundamentals. This is combined with machine learning models which predict the future movements of price and volume for the value stocks which is used to decide whether to trade long or short. Different optimization techniques such as mean-variance optimization and minimax optimization are studied and tested on the trading strategies to allocate optimal weights to the underlying assets in the portfolios. These strategies are developed using python programming language and using the Quantopian API for data collection and back-testing in a trading environment.