Portfolio Management Using Artificial Trading Systems Based on Technical Analysis

摘要

Evolutionary algorithms consist of several heuristics able to solve optimization tasks by\udimitating some aspects of natural evolution. In the field of computational finance, this type of\udprocedures, combined with neural networks, swarm intelligence, fuzzy systems and machine\udlearning has been successfully applied to a variety of problems, such as the prediction of stock\udprice movements and the optimal allocation of funds in a portfolio.\udNowadays, there is an increasing interest among computer scientists to solve these issues\udconcurrently by defining automatic trading strategies based on artificial expert systems,\udtechnical analysis and fundamental and economic information. The objective is to develop\udprocedures able, from one hand, to mimic the practitioners behavior and, from the other, to\udbeat the market. In this sense, Fernandez-Rodríguez et al. (2005) investigate the profitability\udof the generalized moving average trading rule for the General Index of Madrid Stock Market\udby optimizing parameter values with a genetic algorithm. They conclude that the optimized\udtrading rules are superior to a risk-adjusted buy-and-hold strategy if the transaction costs\udare reasonable. Similarly, Papadamou & Stephanides (2007) present the GATradeTool, a\udparameter optimization tool based on genetic algorithms for technical trading rules. In the\uddescription of this software, they compare it with other commonly used, non-adaptive tools in\udterms of stability of the returns and computational costs. Results of the tests on the historical\uddata of a UBS fund show that GATradeTool outperforms the other tools. Fernández-Blanco\udet al. (2008) propose to use the moving average convergence divergence technical indicator\udto predict stock indices by optimizing its parameters with a genetic algorithm. Experimental\udresults for the Dow Jones Industrial Average index confirm the capability of evolutionary\udalgorithms to improve technical indicators with respect to the classical configurations adopted\udby practitioners.\udAn alternative approach to generate technical trading systems for stock timing that combines\udmachine learning paradigms and a variable length string multi-objective genetic algorithm\udis proposed in Kaucic (2010). The most informative technical indicators are selected by\udthe genetic algorithm and combined into a unique trading signal by a learning method. A\udstatic single-position automated day trading strategy between the S&P 500 Composite Index\udand the 3-months Treasury Bill is analyzed in three market phases, up-trend, down-trend\udand sideways-movements, covering the period 2000-2006. The results indicate that the near-optimal set of rules varies among market phases but presents stable results and is able to\udreduce or eliminate losses in down-trend periods.\udAs a natural consequence of these studies, evolutionary algorithms may constitute a\udpromising tool also for portfolio strategies involving more than two stocks. In the field of\udportfolio selection, Markowitz and Sharpe models are frequently used as a task for genetic\udalgorithm optimization. For instance, the problem of finding the efficient frontier associated\udwith the standard mean-variance portfolio is tackled by Chang et al. (2000). They extend\udthe standard model to include cardinality and composition constraints by applying three\udheuristic algorithms based upon genetic algorithms, tabu search and simulated annealing.\udComputational results are presented for five data sets involving up to 225 assets.\udWilding (2003) proposes a hybrid procedure for portfolio management based on factor\udmodels, allowing constraints on the number of trades and securities. A genetic algorithm\udis responsible for selecting the best subset of securities that appears in the final solution, while\uda quadratic programming routine determines the utility value for that subset. Experiments\udshow the ability of this approach to generate portfolios highly able to track an index.\udThe β − G genetic portfolio algorithm proposed by Oh et al. (2006) selects stocks based on\udtheir market capitalization and optimizes their weights in terms of portfolio β’s standard\uddeviation. The performance of this procedure depends on market volatility and tends to\udregister outstanding performance for short-term applications.\udThe approach I consider for portfolio management is quite different from the previous models\udand is based on technical analysis. In general, portfolio optimizations using technical analysis\udare modular procedures where a module employs a set of rules based on technical indicators\udin order to classify the assets in the market, while another module concentrates on generating\udand managing portfolio over time (for a detailed presentation of the subject, the interested\udreader may refer to Jasemi et al. (2011)). An interesting application in this context is the approach developed by Korczak & Lipinski\ud(2003) that leads to the optimization of portfolio structures by making use of artificial trading\udexperts, previously discovered by a genetic algorithm (see Korczak & Roger (2002)), and\udevolutionary strategies. The approach has been tested using data from the Paris Stock\udExchange. The profits obtained by this algorithm are higher than those of the buy-and-hold\udstrategy.\udRecently, Ghandar et al. (2009) describe a two-modules interacting procedure where a genetic\udalgorithm optimizes a set of fuzzy technical trading rules according to market conditions and\udinteracts with a portfolio strategy based on stock ranking and cardinality constraints. They\udintroduce several performance metrics to compare their portfolios with the Australian Stock\udExchange index, showing greater returns and lower volatility.\udAn alternative multi-modular approach has been developed by Gorgulho et al. (2011) that\udaims to manage a financial portfolio by using technical analysis indicators optimized by a\udgenetic algorithm. In order to validate the solutions, authors compare the designed strategy\udagainst the market itself, the buy-and-hold and a purely random strategy, under distinct\udmarket conditions. The results are promising since the approach outperforms the competitors.\udAs the previous examples demonstrate, the technical module occupies, in general, a\udsubordinate position relative to the management component. Since transaction costs, cardinality and composition constraints are of primary importance for the rebalancing\udpurpose, the effective impact of technical signals in the development of optimal portfolios\udis not clear. To highlight the benefits of using technical analysis in portfolio management,\udI propose an alternative genetic optimization heuristic, based on an equally weighted zero\udinvestment strategy, where funds are equally divided among the stocks of a long portfolio\udand the stocks of a short one. Doing so, the trading signals directly influence the portfolio\udconstruction. Moreover, I implement three types of portfolio generation models according to\udthe risk-adjusted measure considered as the objective, in order to study the relation between\udportfolio risk and market condition changes.\udThe remainder of the chapter is organized as follows. Section 2 explains in detail the proposed\udmethod, focusing on the investment strategy, the definitions of the technical indicators and\udthe evolutionary learning algorithm adopted. Section 3 presents the experimental results and\uddiscussions. Finally, Section 4 concludes the chapter with some remarks and ideas for future\udimprovements.