اتخاذ الگوریتم های ژنتیکی برای تجزیه و تحلیل فنی و مدیریت پرتفولیو

This research examines two different applications of the Genetic Algorithms (GA) in portfolio management. GA is adopted to determine the optimized parameters setting of different technical indicators and portfolio weighting. Besides the Traditional GA, the Hierarchical GA is also adopted in this research. Different algorithms and the usage of different numbers of technical indicators are evaluated in different economic situations. GA shows its optimization power over different tasks in portfolio management.

The number of people who participate in the stock exchange market grows rapidly. This generates the needs of developing the tools for the investors to manage their investment portfolios [1]. It is difficult for the investors to well distribute their stocks on hand. Wrong allocation of investment affects the return of their investment. However, if the investment is allocated correctly, the portfolio will contain less risk and higher return. Therefore, proper investment distribution and portfolio management are required. On the other hand, technical analysis [2] is a security analysis methodology for forecasting the future direction of security prices through the study of the past market data, the primarily price and the volume. Technical analysts may employ models and trading rules especially based on price and volume transformations, such as the Relative Strength Index (RSI), and Moving Averages (MA), etc. Each technical indicator has its own characteristic. However, it is difficult to set the corresponding parameters of different indicators because of their unique behavior. Therefore, an optimization strategy for tuning the parameters of different indicators is necessary. The objectives of this research are to tackle the problems of the two areas mentioned above, i.e. (1) the difficulty of allocating the portfolio weighting based on the Genetic Algorithms (GA) and (2) the ambiguity in setting the parameters of the technical indicators. A brief review on adopting GA to portfolio management will be given in Section 2. Section 3 focuses on the second objective, GA is proposed to determine the parameters of different technical indicators. Applying GA for portfolio weighting determination will be introduced in Section 4. The experimental result will be discussed in Section 5 whereas the conclusion will be made in Section 6.

In this research, GA is adopted to determine the optimum parameters settings of different technical indicators. After the portfolio is formed, GA is adopted again to obtain the distribution of stock weighting in the portfolio. This totally affects the risk, the return and the Sharpe ratio of the portfolio. Besides the Traditional GA, the Hierarchical GA is also introduced to improve the performance. The experimental result shows that the Hierarchical GA always has the highest standard deviation in the fitness. This is caused by the Elite Population which has a lower crossover and mutation rates. However, it takes more generations to achieve convergence. The Traditional GA achieves faster convergence, but its variation is relatively lower which causes the highest fitness fluctuates from generation to generation. When tuning the parameters of different technical indicators, the GA optimization performs well in both Bearish and Bullish markets. The result also shows that using multiple indicators is better than a single indicator. However, the Buy and Hold strategy outperforms the GA optimization in the Bullish market. Moreover, the only profit making investment of the Buy and Hold strategy in Bearish market is in the Utilities sectors. This also verifies the saying that “Invest in the Utilities sector while there is a Bearish market”. In addition, this research also investigates how different weighting strategies affect a portfolio’s performance. From the equal weighting strategy, the selection of stock becomes the critical criterion that affects the performance of a portfolio. This strategy reflects the individual stock’s performance directly. However, the risk and the return as well as the Sharpe Ratio are relatively poorer than other strategies. In the Markowitz weighting strategy, the weighting of the stocks is based on the efficient frontier with tangent to the transformation line. While using the proposed GA weighting strategy, the weighting on stock is optimized by the Sharpe Ratio. It gives the best performance in the experiment which means that the portfolio determined has a lower risk and a higher return when comparing in the unit of risk to the return ratio. It also demonstrates the scenario that “Purchase Utilities stocks when there is a Bearish market”. To sum up, GA is adopted to determine the optimized value of the parameters of different technical indicators in this research. Optimized technical indicators provide a direction for buying and selling of stocks in any market condition. The result also shows that the combination of multiple indicators outperforms the single one. The stock allocation in a portfolio can then be optimized by GA as well. For future work, a real time optimization strategy will be proposed to reflect the rapid market change.