Balancing a stock portfolio

• Main Author: Chong, Isabelle Noelene
• Other Authors: Rajapakse Jagath Chandana
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/62894
• Institution: Nanyang Technological University

Attaining an optimal stock portfolio is the one of the main goals for most investors. In order to do so, they find methods to help them in identifying the profitable stocks, predicting the market movements and optimizing their investment portfolio. However, this optimal portfolio will not be able to last through time. Hence, a more long-term goal would to be able to rebalance that portfolio whenever necessary. Rebalancing is the process of realigning the weightings of one's portfolio of assets. It involves buying or selling assets in your portfolio to maintain your original desired level of asset allocation. This is necessary as prices of stocks are frequently changing, which would result in the optimal allocation of stocks in a portfolio changing as well. This project involves designing an intelligent system for portfolio replication, optimization as well as rebalancing based on past performance of the investments over a specified period. Artificial neural networks (ANN) and linear regression will be used for portfolio replication; quadratic programming will be used for portfolio optimization and portfolio rebalancing will be modelled using the control-theoretical model. The objective functions used in portfolio optimization are largely based on the Modern Portfolio Theory (MPT) by Harry Markowitz. Firstly, for portfolio replication, both ANN and linear regression approaches were shown to be able to replicate the Straits Times Index (STI) portfolio with high correlation values. Secondly, for portfolio optimization, we explore three different types of portfolio configurationsmean-variance optimized portfolio, mean-variance optimized portfolio without shorting, as well as mean-variance optimized portfolio without shorting and heavy concentration. Quadratic programming is able to search for an optimized portfolio according to the investor’s risk appetite and investment criterions. Lastly, for portfolio rebalancing, we will be using the control-theoretical model to apply rebalancing for different monitoring frequencies by using periodic rebalancing, different thresholds using threshold rebalancing as well as rebalancing to different extent of the target allocation. The performances of the portfolio using different rebalancing techniques will then be compared using a set of performance metrics. The dataset used for the experiments above are June 2012 to June 2013 (1 year period) as well as an extended period of June 2012 to June 2014 (2 year period). R software programming language is used for the implementations and experiments for the entire project. The experiments are conducted with the historical price data extracted from Yahoo! Finance.