قوانین فازی بر پایه سیستم های خبره برای ارزیابی سهام و ساخت و ساز پرتفولیو : به درخواست بورس اوراق بهادار استانبول
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|22007||2013||13 صفحه PDF||سفارش دهید||8920 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 40, Issue 3, 15 February 2013, Pages 908–920
The aim of this study is to construct appropriate portfolios by taking investor’s preferences and risk profile into account in a realistic, flexible and practical manner. In this concern, a fuzzy rule based expert system is developed to support portfolio managers in their middle term investment decisions. The proposed expert system is validated by using the data of 61 stocks that publicly traded in Istanbul Stock Exchange National-100 Index from the years 2002 through 2010. The performance of the proposed system is analyzed in comparison with the benchmark index, Istanbul Stock Exchange National-30 Index, in terms of different risk profiles and investment period lengths. The results reveal that the performance of the proposed expert system is superior relative to the benchmark index in most cases. Additionally, in parallel to our expectations, the performance of the expert system is relatively higher in case of risk-averse investor profile and middle term investment period than the performance observed in the other cases.
Portfolio management process is an integrated set of steps undertaken in a consistent manner to create and maintain an appropriate portfolio to meet clients’ goals (Maginn, Tuttle, McLeavey, & Pinto, 2007). The aim of this study is constructing appropriate portfolios that meet investor’s risk profile and specific preferences, rather than constructing an optimal portfolio that is just a collection of individual assets having desirable risk-return characteristics. Accordingly, a fuzzy rule based expert system (ES) is developed in this study to support portfolio managers in their middle term stock evaluation and portfolio construction decisions. Modern Portfolio Theory (MPT), as an important research area of modern finance theory, has born from the study of Markowitz published in 1952. Markowitz (1952) showed that the variance of the rate of return was a meaningful measure of portfolio risk under a set of assumptions, and he derived the formula for computing the variance of a portfolio. MPT has been widely accepted and studied by researchers. However, in recent years, criticism on the assumptions of MPT is increasing. The basic assumption of MPT is the efficiency of markets. However, Grossman and Stiglitz (1980) asserted that obtaining information about markets is costly and it is impossible to get whole information about each individual stock. Therefore, prices cannot perfectly reflect the information and markets cannot be efficient. Hence, it is very important to identify the undervalued stocks for investment. Another criticism on MPT is the computational burden caused by the quadratic utility functions and covariance matrix. This burden causes challenging difficulties in real life applications due to the high number of stocks. That is why investors prefer to use simplified investment rules instead of the models in the field of MPT. However, the portfolio management process is divided into two stages in recent studies to reduce the initial number of stocks and consequently reduce the computational difficulty. In the first stage, appropriate stocks for portfolio construction are selected. In the second stage, the amount of capital to be invested in each selected stock is specified. The study of Xidonas, Askounis, and Psarras (2009) is an example to this two-stage process. Finally, it is widely criticized that MPT disregards real investor’s preferences. Moreover, it is often found in portfolio optimization that investors prefer portfolios that lie behind the efficient frontier of the Markowitz model even though they are dominated by other portfolios with respect to the two criteria, expected return and risk. This observation can be explained by the fact that not all the relevant information for an investment decision can be captured in terms of explicit return and risk. Therefore, some additional criteria must be added to the classical risk-return criteria. By considering additional and/or alternative decision criteria, a portfolio that is dominated with respect to expected return and risk may make up for the deficit in these two criteria by a very good performance in one or several other criteria and thus be non-dominated in a multi-criteria setting (Ehrgott, Klamroth, & Schwehm, 2004). As a result, portfolio management is a multidimensional problem and multi-criteria decision making (MCDM) approach provides the methodological basis to resolve the inherent multi-criteria nature of the problem. MCDM approach builds realistic models by taking into account, apart from the two basic criteria; return and risk, a number of important other criteria, i.e. additional statistical measures of the variation of return, criteria that are founded in the theory of fundamental analysis, or criteria related to the stock market characteristics and behavior of securities, etc. (Xidonas, Mavrotas, Zopounidis, & Psarras, 2011). Additionally, MCDM have the advantage of taking into account the preferences of any particular investor. Furthermore, these methods do not impose any norm to the investor’s behavior. The use of MCDM methods allows synthesizing in a single procedure the theoretical and practical aspects of portfolio management, and then it allows a non-normative use of theory (Xidonas & Psarras, 2009). Portfolio management is a complex, subjective and generally unstructured process. Additionally, decision makers have partial information about the market and have to deal with high level of uncertainty. Moreover, the interaction between fundamental and technical criteria is uncertain. Due to the complex, uncertain and unstructured nature of the problem, there is a growing interest in artificial intelligence (AI) techniques recently. The readers who are interested in more details regarding AI applications on portfolio management may refer to Bahrammirzaee (2010). Among these techniques, a fuzzy rule based ES is thought to be an appropriate framework for the solution due to the characteristics of the problem. In this study, a fuzzy rule based expert system is developed to support portfolio managers in their middle term investment decisions. The proposed expert system is validated by using the data of Istanbul Stock Exchange (ISE) National-100 Index (XU100). The remainder of this study is organized as follows: In Section 2, previous studies in which ES technique is used to solve stock evaluation and portfolio construction problem are presented. In Section 3, structure of the proposed ES is explained. In Section 4, performance of the proposed ES is analyzed by using the historical data obtained from ISE in cases of different risk profiles and investment period lengths. In Section 5, the performance evaluation results are discussed. Finally, concluding remarks are presented in Section 6.
نتیجه گیری انگلیسی
A fuzzy rule based ES is developed in this study to support portfolio managers in their middle term stock evaluation and portfolio construction decisions. The aim of this study is to construct an appropriate portfolio that meets investor’s risk profile and specific preferences, rather than constructing an optimal portfolio that is just a collection of individual assets having desirable risk-return characteristics. The proposed ES can be characterized by its realistic, flexible and practical aspects. As the proposed ES uses relative FRs and calculates the data ranges dynamically, in fuzzification of inputs, it can be employed in solving real-life problems. Additionally, the proposed ES is flexible, since it can be tailored according to the investor’s risk profile and specific preferences by changing some parameters simply. Moreover, the proposed system is practical, as users can easily understand its structure and they can adjust its parameters conveniently according to their preferences. Furthermore, as stocks can be evaluated through a single process by using relative FRs, implementation of the proposed ES is convenient. The proposed ES is validated by using the data on 61 stocks that publicly traded in XU100 from the years 2002 through 2010. The performance of the proposed ES is analyzed in comparison with the benchmark index, XU030, in terms of different risk profiles and investment period lengths. The results reveal that the proposed ES outperforms the benchmark index in terms of all risk profiles. More specifically, it performs relatively better in the risk averse investor case than it does in the other cases. Additionally, the performance of the proposed ES is superior relative to the benchmark index in terms of different investment period lengths. More specifically, the performance of the system better in the middle term investment periods. In parallel to our expectations, the performance of the proposed ES is relatively higher in risk averse investor and middle term investment period cases.