Elsevier

Applied Soft Computing

Volume 37, December 2015, Pages 142-155
Applied Soft Computing

Combined soft computing model for value stock selection based on fundamental analysis

https://doi.org/10.1016/j.asoc.2015.07.030Get rights and content

Highlights

  • A combined soft computing model for value investing is proposed.

  • The DRSA model generated 20 rules to classify value stocks.

  • Two strong decision rules were obtained for conducting FCA analysis.

  • The selected “Good” value stock portfolio outperformed the market index.

  • Implications for value stock selection are obtained by DRSA and FCA.

Abstract

The stock selection problem is one of the major issues in the investment industry, which is mainly solved by analyzing financial ratios. However, considering the complexity and imprecise patterns of the stock market, obvious and easy-to-understand investment rules, based on fundamental analysis, are difficult to obtain. Therefore, in this paper, we propose a combined soft computing model for tackling the value stock selection problem, which includes dominance-based rough set approach, formal concept analysis, and decision-making trial and evaluation laboratory technique. The objectives of the proposed approach are to (1) obtain easy-to-understand decision rules, (2) identify the core attributes that may distinguish value stocks, (3) explore the cause–effect relationships among the attributes or criteria in the strong decision rules to gain more insights. To examine and illustrate the proposed model, this study used a group of IT stocks in Taiwan as an empirical case. The findings contribute to the in-depth understanding of the value stock selection problem in practice.

Introduction

In this paper, we propose a combined soft computing model for value stock selection that is based on financial ratio analysis, also termed “fundamental analysis” (FA). The stock selection problem can be traced back to the efficient market hypothesis (EMH) [1], which assumes that investors cannot use available information to form an investing strategy for consistently outperforming the stock market. In the hope of finding a useful strategy, researchers have used various investment plans to examine the EMH [2], [3], [4]. Among those investing strategies, value investing is widely examined in academia and in practice [5]; it originated from the classical work of Graham and Dodd [6]. Value investing strategy is based on the idea that out-of-favor stocks are sometimes underpriced because of the inefficiency of the stock market. Smart investors may earn extra premium by investing in these underpriced stocks. In finance, researchers often classify stocks that have high book-to-market equity (B/M) or earnings-to-price (E/P) ratios as value stocks [7]; the value premium has been discovered in many markets [2], [3], [7], [8], [9]. Though the essential idea of value investing is generally accepted, further selection among a group of generic high B/M or E/P ratio stocks is a challenging task. In practice, the evaluation of the worthiness and prospect of a value stock mainly relies on FA [10], [11]. Although the use of relevant financial variables to assess the prospect of a stock is widely adopted, opinions on the inclusion of information and analyzing methods are divided.

The stock selection problem is solved using two main approaches. In the conventional approach, financial studies tend to use regression models for determining the relationship among historical financial ratios and future earnings (or stock performance). To examine the usefulness of a value investing strategy, Piotroski [9] included nine financial variables from three aspects to discriminate winners and losers by forming a logit-regression model (F-Score model). Similarly, Mohanram [12] developed a G-Score model for selecting glamor stocks. Both the F- and G-Score models obtained positive results in their proposed experimental periods. The aforementioned studies mainly used regression models, which might be suitable for explaining some phenomenon; nevertheless, the complexity and nonlinear relationships among financial ratios and the subsequent stock returns often impede practitioners in determining useful investing rules in a specific context [13].

In addition, in the nonconventional approach, researchers from other fields, such as artificial intelligence (AI) and multiple criteria decision making (MCDM) have attempted to leverage the computational strength of computer programming to solve the complex stock selection problem. Among various AI techniques, artificial neural networks (ANNs) are widely used to predict stock performance because of their superior capability to model a nonlinear relationship. For instance, Lam [10] combined the fundamental and technical variables in the ANN model to predict financial performance. Recent studies have proposed an integration approach for enhancing the learning capability of the ANN model. For example, Hadavandi et al. [14] integrated the genetic fuzzy system with ANN to forecast stock prices. Hsu [15] used a hybrid approach to combine ANN with genetic programming. Although the aforementioned studies generated positive outcomes from their models, drawing useful conclusions for future investment from the learning results of ANNs is often difficult. The black-box processing characteristic of ANN and support vector machine (SVM) [16] impedes investors in understanding the complex relationship among the considered variables of each model. In another nonconventional approach, MCDM considers multiple criteria for tackling the complex stock rankings and selection problem. The group decision methods—analytic network process (ANP) and decision-making trial and evaluation laboratory (DEMATEL)—were used for retrieving experts’ domain knowledge regarding investment practice [13], [17]. The obtained results hinged on the experts’ subjective judgments. Each experts provided his opinions (by questionnaire) to incorporate his knowledge into the model. Another main method in the MCDM is the data envelope analysis (DEA) that uses mathematical programming to gauge the efficiency of the considered stocks [18], [19]. Although the DEA method is useful for portfolio selection, it cannot explore the relationships among financial variables.

As discussed, existing studies (i.e., statistics and certain AI techniques) have limitations in exploring the complex relationships among financial variables to induct applicable knowledge of value stocks. Therefore, the major motivation of this study was to devise a model that may relax the unrealistic assumptions of statistics and tackle the vague and imprecise financial data to obtain understandable knowledge or implications for value stock selection. As a result, this paper proposes a combined soft computing model by using the dominance-based rough set approach (DRSA), formal concept analysis (FCA), and DEMATEL techniques, to obtain applicable knowledge for this problem. The conceptual framework of this study is illustrated in Fig. 1; in addition, the purpose and advantages of the proposed approach (compared with previous techniques, such as statistics, ANNs, and SVM) are highlighted.

The key parts of the proposed approach are discussed as below. First, the DRSA classifier reduces the considered criteria (i.e., obtained CORE criteria) and generates a set of decision rules to select value stocks with strong prospects. Compared with the AI techniques (e.g., ANNs and SVM), DRSA may generate a set of decision rules for investors to comprehend (instead of a black-box). In addition, DRSA may consider partial sets of criteria in each context (i.e., decision rule), which is closer to how domain experts make judgments (ANNs, DEA, or SVM must process all input variables simultaneously to adjust their parameters during the learning phase); it is difficult for a human brain to make decisions based on all criteria in each time [20]. In addition, DRSA does not have to assume the distribution of data or the independence of variables. Second, the FCA is incorporated to induct the plausible symptoms that may satisfy the premises in the DRSA decision rules. This deepens the understanding of decision rules obtained from objectively inducted implications, which may enhance the findings from DRSA. Furthermore, the CORE criteria from the DRSA model could be analyzed using the DEMATEL technique, which may leverage domain experts’ knowledge to identify cause–effect relationships (directional influences) among the CORE dimensions (criteria). Last, the obtained symptoms and source criteria of decision rules from FCA and DEMATEL analyses may complement each other to support the result. These findings along with the strong DRSA decision rules, could be illustrated as a directional flow graph (DFG) for enhancing the understanding of value stock investment in practice. The proposed approach not only supports the identification of value stocks with superior prospects, but also aims at retrieving understandable implications for investors, which have been underexplored in previous studies.

To illustrate the proposed approach, Taiwan's stock market was examined as an empirical case. The stock market was regarded as a “big data set,” composed of more than 1400 stocks and numerous financial attributes. Inducting effective and critical patterns from the big data set is a challenging yet valuable task. The proposed soft computing model intends to explore or retrieve the effective patterns from the stock market with easy-to-understand decision rules for supporting the value stock selection problem.

The remainder of this paper is divided into five sections. Section 2 gives a short review of FA and discusses the proposed DRSA, FCA, and DEMATEL techniques. Section 3 introduces the DRSA methodology and the FCA. In Section 4 the proposed method is used to explore Taiwan's stock market as an empirical case with an experimental result. Section 5 discusses the result, and Section 6 concludes this paper.

Section snippets

Preliminary

This section reviews studies related to value investing strategy and FA. In addition, the proposed DRSA, FCA, and DEMATEL techniques are briefly discussed. Furthermore, the reason for using the DRSA method for solving the value stock selection problem is explained.

Methodology

The aim of the proposed methodology is to discover useful knowledge for selecting value stocks. As the evaluation of a stock comprises multiple dimensions and pertaining criteria, it is reasonable to adopt the MCDM approach for the analyses. In this study, we proposed the DRSA methodology as a knowledge discovery system at the initial stage to explore applicable rules for the value investing strategy. The conceptual framework of the proposed model is illustrated in Fig. 1, and the combining of

Empirical case and research results

A group of publicly listed stocks was examined for illustrating the proposed model. Because Taiwan plays a major role in the global IT industry, the IT sector is not only crucial to the economic growth of Taiwan but also supports global advancements in technology; therefore, we selected the publicly listed stocks from the IT industry in Taiwan as an empirical case. The experiment processes and their purposes are illustrated in Fig. 2.

Discussion

The experimental results suggest that the proposed DRSA model could distinguish value stocks with satisfactory financial returns in the subsequent period, which included multiple financial attributes and 20 decision rules (i.e., 20 = 9 + 11; nine decision rules associated with the “Good” decision class and 11 associated with the “Bad” decision class). A group of generic high B/M ratio stocks (74 stocks) from the IT sector in 2011 were used to obtain the DRSA model, which was examined by a group of

Conclusions

This study used the combined soft computing model to explore the usefulness of FA in selecting satisfactory value stocks. The three major contributions of this study are summarized as follows: (1) Identify the contexts (decision rules) and core criteria to discern value stocks with satisfactory prospects; (2) Explore the cause–effect relationships among the CORE attributes (by using the DEMATEL technique), which enables investors to observe the source factors (from FCA) that might improve or

References (53)

  • J.J.H. Liou

    Variable consistency dominance-based rough set approach to formulate airline service strategies

    Appl. Soft Comput.

    (2011)
  • J.J.H. Liou et al.

    Using decision rules to achieve mass customization of airline services

    Eur. J. Operat. Res.

    (2010)
  • S. Greco et al.

    Rough sets methodology for sorting problems in presence of multiple attributes and criteria

    Eur. J. Operat. Res.

    (2002)
  • J. Błaszczyński et al.

    Multi-criteria classification—a new scheme for application of dominance-based decision rules

    Eur. J. Operat. Res.

    (2007)
  • J. Poelmans et al.

    Formal concept analysis in knowledge processing: a survey on applications

    Expert Syst. Appl.

    (2013)
  • S.K. Fang et al.

    Exploring the preference of customers between financial companies and agents based on TCA

    Knowl. Based Syst.

    (2012)
  • C.H. Hsu et al.

    The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR, resources

    Conserv. Recycl.

    (2012)
  • S.K. Hu et al.

    Exploring smart phone improvements based on a hybrid MCDM model

    Expert Syst. Appl.

    (2014)
  • E.F. Fama et al.

    Permanent and temporary components of stock prices

    J. Polit. Econ.

    (1988)
  • C. Capaul et al.

    International value and growth stock returns

    Financ. Anal. J.

    (1993)
  • L.K.C. Chan et al.

    Value and growth investing: review and update

    Financ. Anal. J.

    (2004)
  • K. Froot et al.

    Style investing and institutional investors

    J. Financ. Quant. Anal.

    (2008)
  • J.D. Piotroski et al.

    Identifying expectation errors in value/glamour strategies: a fundamental analysis approach

    Rev. Financ. Stud.

    (2012)
  • B. Graham et al.

    Security Analysis

    (1934)
  • E.F. Fama et al.

    Value versus growth: the international evidence

    J. Finance

    (1998)
  • H. Gulen et al.

    Value versus growth: time-varying expected stock returns

    Financ. Manag.

    (2011)
  • Cited by (46)

    • A comprehensive decision support system for stock investment decisions

      2022, Expert Systems with Applications
      Citation Excerpt :

      Shen and Tzeng (2015) used seventeen indicators of this type. Common categories of fundamental indicators used during the stock selection stage are: Profitability, Leverage, Liquidity, Efficiency, Growth, Solvency, Operational efficiency (Shen & Tzeng, 2015; Yang et al., 2019). Of course, combining fundamental indicators with other factors is common practice for stock selection.

    • Interactive portfolio optimization model based on rough fundamental analysis and rational fuzzy constraints

      2022, Applied Soft Computing
      Citation Excerpt :

      The second aspect considers various factors, other than return and risk, in screening stocks. One of the mainstreams is applying valuation ratios (e.g., earnings or book value per share) [34], and the other is analyzing various financial ratios simultaneously [20,21,35]. Given the complexity of equity evaluation, this approach leverages different prevailing decision-aided methods (e.g., the ANP [9], ELECTRE [10], TOPSIS [11], and PROMETHEE [36]) to evaluate multiple financial ratios, categorized as fundamental analysis.

    • Share price prediction of aerospace relevant companies with recurrent neural networks based on PCA

      2021, Expert Systems with Applications
      Citation Excerpt :

      The 5 feature vectors, such as profitability, growth, liquidity, solvency, and operational efficiency, were used for tackling the value stock analysis of IT stocks in Taiwan. The profitability vector includes ROA, operational gross profit, operational profit, net profit after tax; the growth vector includes net profit after tax growth rate, ROA growth rate, total assets growth rate, revenue growth rate, gross profit growth rate; the liquidity vector includes quick ratio, liquidity ratio, cash ratio; the solvency vector includes debt ratio, interest coverage ratio; and operational efficiency vector includes asset turnover rate, inventory turnover rate, average days for sales (Shen & Tzeng, 2015a). Besides, the rate of return, the number of transactions, gross profit, gross loss, the number of profitable transactions, the number of consecutively profitable transactions, the number of unprofitable consecutive transactions, Sharpe ratio, the average coefficient of volatility, the average rate of return per transaction, the value of the evaluation have been used in the decision system on FOREX market (Korczak et al., 2016).

    • Improving DEA cross-efficiency optimization in portfolio selection

      2021, Expert Systems with Applications
      Citation Excerpt :

      For simplicity in the illustration, we select two inputs and two outputs for each stock from the raw financial data that is available in the annual financial statement. Similar to the literature (Lim et al., 2014; Mashayekhi & Omrani, 2016; Shen & Tzeng, 2015), the two inputs are Leverage (LEV) and Cash ratios and the two outputs are Return on assets (ROA) and Earnings per share (EPS). The Leverage (LEV) ratio is defined as the proportion of a firm’s debts to its total assets.

    View all citing articles on Scopus
    View full text