A multicriteria DSS for a global stock evaluation

Entxetprl~aK1] 'Epeovct / Operational Research. An International Journal. Vol.3, No.3 (2003), pp.281-306 A Multicriteria DSS for a Global Stock Evaluation Georgios D. Samaras, Nikolaos F. Matsatsinis Technical University of Crete Decision Support Systems Laboratory University Campus, 73100, Chania, Greece {nikos, samarasg}~ergasva.tuc.gr Constantin Zopounidis Technical University of Crete Financial Engineering Laboratory University Campus, 73100, Chania, Greece kostas(&dpem.tuc.gr Abstract The paper describes a Multicriteria Decision Support System which aims at presenting a complete and spherical evaluation of the Athens Stock Exchange stocks. The system evaluates the stocks on the basis of three approaches : Fundamental Analysis, Technical Analysis and Stock-Exchange Analysis. The system introduced in this paper utilises Multicriteria Analysis methods and embodies a large volume of relevant information. It is a 'live' system and operates in 'real world conditions' since its data are updated on a daily basis through a multitude of sources. The final output of the system is an overall evaluation of the ASE stocks and a stock ranking with the best stock first and the worst last, in order to be used in supporting investment decision-making. Finally, the system is intended for both institutional and private investors. Keywords : Multicriteria Decision Support Systems; Stock Evaluation; Fundamental Analysis; Technical Analysis; Stock-Exchange Analysis 1. Introduction Stock evaluation constitutes an integral part of the Portfolio Management (PM) process, which comprises, on the one hand, the determination of the most attractive stocks, and on the other hand, the portfolio composition, which is the process that selects the stocks and determines the percentage in which each stock will participate in the portfolio. Stock evaluation, and Portfolio Management in general, should not be conducted on the basis of the investor's personal perceptions and abilities, but it should be based on the development and application of specific and scientifically founded practices and on the use of the potentialities offered by advanced technology. The field of professional portfolio management is an extremely competitive environ- 282 Operational Research. An International Journal / Vol.3, No.3 / September- December 2003 ment; therefore there is an imperative need for the creation of technologically advanced management tools which will be developed in order to reinforce the arsenal of both professional managers and private investors. During the development process of various management tools, special effort has been made to ensure that the operation of those tools will cover the overall spectrum of problems faced by professional managers as well as private investors. Portfolio Management relies on a great number of factors and involves complicated procedures. Successful PM should comprise the following operations : investment profile definition (investment strategy, investment ho- rizon, capital to be invested, etc.), stock evaluation, selection of the most attractive stocks, definition of portfolio dispersion, portfolio composition, portfolio return evaluation, market timing definition, and other decisions related to PM. A methodological basis that can deal with the multicriteria nature of PM problems is Multicriteria Analysis. Afterwards, PM was the subject of research in other fields, such as Artificial Intelligence. Consecutively, intelligent systems emerged and changed the way PM is viewed. Intelligent systems incorporate human knowledge and experience along with artificial intelligence technologies. The integration of intelligent systems with traditional Decision Support Systems has resulted in the new generation ofDSS : Intelligent Decision Support Systems. Considerable progress has been made in this field so far, but there is a lot to be done so that the overall spectrum of management need will be met satisfactorily in the scope of current, complex and internationalised stock exchange actualities. 2. Theoretical Background 2.1 Stock Evaluation and Portfolio Management Stock Evaluation is one of the basic operations of Portfolio Management. Stock Evaluation is a complex multidimensional problem that can be dealt with successfully only as a combination of scientific methodology, on the one hand, and personal experience of the field professionals, on the other. Thus, all advanced tools will be utilised in order to ensure the integration of the know-how that derives from considerable research activity, with the experience that results from the long-time involvement of professionals in the field of Portfolio Management. These tools are Fundamental Analysis, Technical Analysis, Stock-Exchange Analysis and analysis of the current conjuncture. The use of these tools allows a spherical and complete stock evaluation, which is a prerequisite for the further process of successful portfolio management. A fundamental portfolio theory was developed by Markowitz [Markowitz (1952)], who proposed the well-known model of mean-variance. The model determines a set of effective portfolios, in which the anticipated return is related to the u n - G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 283 dertaken risk, while the multiattribute utility theory allows the selection of one portfolio out of this set. Then, equilibrium models were developed; these models are : a) Capital Asset Pricing Model (CAPM), which is further development of the meanvariance model and the multiattribute utility theory, under performance conditions towards risk (beta) [Sharpe (1964)] and b) Arbitrage Pricing Theory (APT), where the stock return common component is expressed, now, through a number of effect factors, each of which is related to a sensitivity index (beta). Another models category is the evaluation models in which the stock price is predicted on the basis of financial and stock-exchange variables, as well as data from the current financial conjuncture. Finally, the market efficiency theory was developed, which has offered considerable applications in the topic of portfolio management. According to this theory, there is such competition among the investors that the stock prices should represent directly the volume of right information in the stock evaluation, in such a way that the market can not be beaten. Therefore, the most appropriate policy is the policy following the market as long as there is a good dispersion in our portfolio. The scientific advances that have taken place since the time of Markowitz's and the other classic models have led to the development of some alternative models. Selection and portfolio management optimisation models constitute one category of alternative models. Some of the most important of these models are: index models, models based on averaging techniques, network models, optimum portfolio (efficient frontier, geometric mean return, stochastic dominance, etc). A detailed review of the above mentioned models is presented in the work paper of Pardalos et al. (1994). 2.2 Multicriteria DSSs and Intelligent Multicriteria DSSs Decision Support Systems (DSS) are interactive computer systems that use models and data to identify and solve low-structure problems, in order to support decisionmakers in the decision making process [Turban and Aronson (2001)]. A system can be identified as DSS if, among others, comprises a model base, support interactively a decision-making process, and is possibly based on a preferential reasoning. A large category of "classic" DSS includes Multicriteria DSS, which are based on Multicriteria Decision Making. The most important multicriteria analysis methods are the following : a) Multiobjective mathematical programming [Colson and De Bruyn (1989); Lee and Chesser (1980)] b) Multi Attribute Utility Theory (MAUT) [Keeney and Raiffa (1976); Keeney (1992)]. c) Outranking methods, represented by the ELECTRE methods family [Roy (1968), (1978); Roy and Bertier (1971); Yu (1992); Vincke (1992)]. 284 Operational Research. An International Journal / Vol.3, No.3 / September- December 2003 d) Ordinal regression based on preference disaggregation approach, represented by the UTA (UTility Additive) methods family [Jacquet-Lagreze and Siskos (1982)1, UTASTAR [Siskos and Yannakopoulos (1985)], UTA // [Siskos (1980)], UTADIS and its variations [Devaud et al. (1980) ; Zopounidis and Doumpos (1999)1, Quasi-UTA [Beuthe et al. (2000)]. Some interactive multicriteria DSS are : ADELAIS [Siskos and Despotis (1989)], MINORA [Siskos et al. (1993)], MARKEX [Siskos and Matsatsinis (1993)], UTA+ [Kostowski (1996)], FINCLAS [Zopounidis and Doumpos (1998)1, MUSA [Siskos et al. (1998)], MIIDAS [Siskos et al. (1999)], PREFDIS [Zopounidis and Doumpos (2000a)], INVESTOR [Zopounidis and Doumpos (2000b)]. An important category of Multicriteria DSSs is the Intelligent Multicriteria DSSs. They are systems which utilise multicriteria methodologies, on the one hand, and on the other hand at least one AI technology: FINEVA [Zopounidis et al. (1996)1, 1NVEX [Vranes, et al. (1996)], INTELLIGENT1NVESTOR [Samaras and Matsatsinis (20o3)1. 3. Intelligent and Multicriteria DSSs in Stock Evaluation and Portfolio Management 3.1 Multicriteria D S S s in Stock Evaluation and Portfolio Management Multicriteria DSS constitute an important category of DSS with considerable applications in the Portfolio Management field. Multicriteria decision analysis provides the methodological framework, required to accommodate the multicriteria nature of the portfolio management problem. In addition, it leads to the development of realistic models which, besides the two basic criteria of retum and risk, also consider other equally important criteria that derive from fundamental and technical analysis, as well as the investor's profile that represents his goals, preferences and policies. Hurson and Zopounidis (1997), examine two types of problems : stock evaluation and portfolio composition. Stocks are ranked according to MINORA [ Siskos and Yannacopoulos (1985)]; Siskos et al. (1993)] method of preference analysis, while they are classified into categories according to ELECTRE TRI [Yu (1992)] outranking method. Portfolio composition is attained by ADELAIS [Siskos and Despotis (1989)] interactive method of multiple objectives linear programming. Another paper is the portfolio selection using ADELA1S [Zopounidis et al.(1998)], which was developed and applied to a set of fifty-two stocks of the Athens Stock Exchange, for the two-year period of 1989-1990. INVESTOR [Zopounidis and Doumpos (2000b)], is a multicriteria DSS which deals with the problem of portfolio selection and composition. FINCLASis a multicriteria DSS, too, for financial classification problems [Zopounidis and Doumpos (1998)]. Finally, Spronk and Hallerbach [Spronk and Hallerbach (1997)] pro- G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 285 posed a system for supporting individual financial decision making. The system is based on multicriteria analysis. The general framework for portfolio management is decision oriented. It is a very general framework in the sense that it can accommodate any type of investor. The framework is also very specific because it gives room to different settings of the portfolio management problem. 3.2 Intelligent Multicriteria DSSs in Stock Evaluation and Portfolio Management FINEVA [Zopounidis et al. (1996)] system constitutes an Intelligent Multicriteria DSS that combines multivariable statistical analysis, multicriteria analysis and ES technology. INVEX [Vranes, et al. (1996)] is an investment program selection system. Its main characteristic is the combination of five different techniques : heuristics, expert system, fuzzy logic, an investor's-behaviour-risk model and the multicriteria method PROMETHE II. Finaly, INTELLIGENTINVESTOR [Samaras and Matsatsinis (2003)] is an intelligent multicriteria DSS which aims at offering an overall consideration of the PM problem. The system incorporates all the advanced PM tools, such as fundamental analysis, technical analysis, market psychology, and uses both multicriteria analysis methods and AI (expert system) technologies. 4. The proposed System The proposed system is intended to be used as a subsystem within an integrated Portfolio Management system. However, it can be used, as well, as an autonomous investment decision support system. The system utilises methodologies which are considered as the most advanced up-to-date tools for stock evaluation, such as Fundamental Analysis, Technical Analysis and Market Psychology. Each one of the above mentioned tools need an amount of quantitative and qualitative data in order to operate successfully, as well as the 'knowledge' for the effective management of those data. One of the basic features of the proposed system is its ability to use data and knowledge of the "real world", a fact which guarantees the advanced reliability of the system. The proposed system, shown in "Figure 1" uses data bases on the fundamental and technical analysis of the Athens Stock Exchange (ASE). Therefore, it is intended to provide support for investment decisions regarding ASE stocks. However, it is fully parameterised and can be used in other stock exchanges, too, provided it is equipped with the respective data bases. Finally, the proposed system addresses both institutional and private investors. The Stock Rankinkg Multicriteria DSS comprises 4 subsystems : 286 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 9 Fundamental Analysis Multicriteria System 9 Technical Analysis Mukicriteria System 9 Stock-Exchange Analysis Multicriteria System 9 Global Ranking Multicriteria System The output of each of the first three subsystems is a stock ranking based on different criteria for each subsystem. The output of the first three subsystems, along with other factor considerations, forms the input for the fourth subsystem which draws up one overall ranking list. The set of stocks to be invested in will be selected from this ranking list. 4.1 Methodological Approach Stock evaluation is carried out in a three-phase process. The output of each phase is a ranking list on the basis of the respective consideration. Q Phase I : FundamentalAnalysis Company Ranking In the first step, the company is analyzed to determine its financial health. This evaluation is conducted on the Fundamental Analysis (FA) level. FA infer the "position" of the company stock Phase 2 : Technical Analysis Stock Ranking In the second step, the evaluation is conducted stock level. The tool which is used is the Technical Analysis (TA). TA provide the stock "'trend". Phase 3 : Stock-Exchange Analysis Ranking In the third and final step of the evaluation, ratios based on stock-exchange figures, are used. A compromise of these three rankings results in the Global Stock Ranking. G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multieriteria DSS for a global stock evaluation 287 Figure I : MethodologicalFrame of the ProposedMulticriteria DSS 4.2 Stock Evaluation Methodologies Two main methodologies are used for the stock evaluation. They are the most advanced tools for a complete evaluation of stocks : Fundamental Analysis and Technical Analysis. Stock-Exchange Analysis, a third methodology, is added in order to supplement the above methodologies. 2 88 Operational Research. An Intemational Journal / Vol.3, No.3 / September-December 2003 4.2.1 Fundamental Analysis Fundamental Analysis (FA) is the study of sector and company conditions to determine the value of a stock. FA constitutes the basic necessary analysis method of the overall evaluation of a company. The strong and weak points of the company hidden behind the accounting figures are also pinpointed by FA. FA aims to determine the financial health of the company, by useful "ratios '; based on the company's financial statements. FA is one of the most effective analysis methods since it provides a dynamic analysis of the company's financial situation and evaluate with precision the advantages and disadvantages of the company. The outlook of the stock is determined by the above analysis, given that the stock is a picture of the company for shareholders and potential investors, in order to predict the future course of stock prices. Fundamental Analysis requires a series of financial statements (Annual Balance Sheets and Annual Results) of the last 3-5 years, as well as, qualitative information regarding all listed companies. This information will be converted to Financial and Qualitative Ratios. The FA infers the "position" of the company, which represents the "real value" of the stock. Afterwards, the "real value" is compared to the "market value" of the stock, in order for the stock to be rated as "overvalued" or "undervalued". The estimate of each stock position may change every time new data from the company accounting figures enter the respective data base. The update of the data base is conducted on a year, semester or quarter basis, depending on the availability of the data [Weston (1986); Conso (1981); Depallens (1980); Gualino (1979); Vizzanova (1981); Giese (1981)]. The ASE provides the Annual Balances, Annual Results, as well as Quarterly Accounting Statements of all the companies listed on the ASE. These data constitute the Functional Balance, which in turn will form the basis for the financial analysis of an enterprise. These financial statements are made out according to the Accounting Plan (AP) to which each company belongs. There are four Accounting Plans in Greece : 9 Commerce/Industry AP 9 Banks AP 9 Insurance Companies AP 9 Investment Companies AP Further information on Investment Companies is also drawn from the EEX Price Bulletin, which is issued monthly by AGII. A data base of the FA ratios, for every AP category is created based on this information (Appendix 1). 4.2.2 Technical Analysis Technical Analysis (TA) is an evaluation method, which is based on the study of the stock charts and the analysis of the data from the daily stock and indices closing prices, aiming at providing the safest and most accurate prediction of the stock prices. TA predicts the stock prices by reviewing only past stock prices, but not Fundamentals. The philosophy supporting TA is based on the hypothesis that stock prices an- G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 289 ticipate all future events. Likewise, TA frequently analyses the recurring performance of stock prices in order to infer some signs of their future performance. Technical Analysis provides the stock trend. The estimation of each stock trend changes every day since the respective data base is updated on the daily stock closing prices [Achelis (2001); Murphy (1986); Edwards (1992); Nison (1991)]. TA draws information from ASE. On a daily basis it receives, via Internet, the closing prices files, for: 9 Stocks, 9 CI, FTSE/ASE-20, MID-40, SMALLCAP-80, Parallel Market Index. 9 Sector Indices 9 Further information about stocks, sectors, CI, etc. All this information provides data to a data base with the Stock Technical Analysis Indices & Ratios. Obviously, there is no stock ranking according to the Accounting Plan category, since the same Technical Analysis is conducted for all stocks (Appendix 2). 4.2.3 Stock-Exchange Analysis In addition to the first two considerations, a third consideration supplements the other two. It is stock-exchange analysis, which is of major importance for stock evaluation. It provides what both shareholders and investors think of the stocks. Furthermore, it provides useful information on the way the stocks are traded in the stock market sessions. Most investors evaluate, more or less, the stocks based on these stockexchange figures. They are data related to profitability and corporate dividend policy, as well as pure stock-exchange figures, such as the current stock-exchange price, etc. The stock-exchange analysis draws information from : 9 Annual Balance Sheets and Annual Results. These data are updated every three months according to the Quarterly Financial Statements. 9 Daily stock closing prices. The combination of the above data creates a data base with the Stock-Exchange Analysis Ratios. Obviously, there is no stock ranking according to the Accounting Plan, since Stock-Exchange Analysis is applies to all stocks (Appendix 3). 4.2.4 Methodology Selection Fundamental Analysis considers all the financial variants that lead investors to their decisions and predicts only on the basis of each company's and each sector's "real" potentials. It evaluates positively only those companies that present the best prospects of improving their financial figures. It is a reliable method to be used particularly for the long-term investment horizon, and it ignores short-term stock price fluctuation. On the other hand, Technical Analysis saves cost and time of collecting and analyzing the data on which Fundamental Analysis is based, since it considers only charts and stock closing prices. Thus, TA is interested only in the "symptoms" the 290 Operational Research. An International Journal / Vol.3, No.3 / September- December 2003 stock or sector shows without studying the "causes" of those symptoms (Fundamental Analysis does it). An important advantage of Technical Analysis is that it integrates the so-called "inside information" on the market, by detecting in good time the stock course that reflects such inside information and activates 'buy' and 'sell' signals. Technical Analysis aims mainly at predicting within the medium-short term horizon. A system that intends to support investment decisions should consider both the parameters of the problem : 9 On the one hand, it should detect the "symptoms" a stock or a sector shows and should make predictions on the basis of these symptoms. Technical Analysis is the tool for the detection of these symptoms. 9 On the other hand, it should analyse the "causes" that rise the above symptoms, so as to provide a clear picture of the future of a stock. Fundamental Analysis is needed for this analysis. Therefore, the proposed Multicriteria DSS evaluates stocks using both Fundamental Analysis and Technical Analysis. Stock evaluation is completed with StockExchange Analysis, which is most important for an overall stock evaluation. 4.3 Muiticriteria Methodological Background The ranking carried out by the four subsystems is based on the multicriteria ranking method UTASTAR [Siskos and Yannakopoulos (1985)], which is an improved version of the UTA [Jacquet-Lagreze and Siskos (1982)] method, according to which, the problem of ordinal regression is the following : Having a weak-order preference structure (>, ~), with ">" the strict preference and "~" the indifference on a set of alternative actions, adjust additive utility functions based on multiple criteria in such a way that the resulting preference structure would be as consistent possible with initial structure. Reference set is the set of referential alternative actions A={a, b, c ..... k), which can be ranked by the decision-maker in order of preference, according to the relations P and I (preference structure). Criteria gl, g2, .... g,, can be either quantitative or qualitative and are defined as real monotone functions : gi : A---~ [gi,, gi*] c R, where gi, is the least desirable value, and gi* is the most desirable value for the criterion gi. Ifgi(a) is the estimate of alternative a, regarding criterion gi, then vector g(a) = {g1(a), g2(a) . . . . . gn(a)}, represents the profile of decision a. The utility function, under certainty, is a real function defined as follows : u : H [gi*, gi*]---> R, for which the following attributes are true : A P b r u[g(a)] > u[g(b)] A I b r u[g(a)] = u[g(b)] ga, b e A G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 291 Criteria Aggregation Model Given the initial preference structure of the alternative actions in the reference set, by the decision-maker, UTA method is designed to estimate a system of utility functions, according to the additive model : U(g) = u(gO + u(gz) + ... + u(g,) subject to the set of regularisation constraints: ui(gi.) = O, ui(gi) 2 O, i = 1,2, .... n ul(gl *) + u2(g2 *) + ... +un(g,*) = 1 In the initial version of UTA method there is a unique error function tr : A [0,1], where c(a) is the amount of utility that must be added in the estimated utility U[g(a)] of the alternative a, in order to make it possible for this action, to regain its rank in the weak-order However, this error function is not sufficient to completely minimise the dispersion of points all around the monotone curve (under- and overestimation errors). In UTA*, is used a double positive error function, which allows to better stabilize the position of the points around the curve. Thus, the global utility function of an alternative action will have the following form : U[g(a)] + a+(a) - #(a), where tr§ and a(a), represent the over-estimation and under-estimation errors of alternative a respectively. UTA* algorithm includes two phases. In the first phase, there is an estimation of the optimal marginal utilities and the global utility of every alternative action a that belongs to A, according to the additive utility model. In the second phase, there is a stability analysis of the global optimum. In case of non uniqueness, the optimal solutions that maximise the criteria weight are determined. For the estimation of marginal utilities in the first phase, and the stability analysis of the global optimum in the second phase, linear programming formulation is used, aiming at finding the optimal utility functions that minimise a global error function. k M i n F = E [ c r + ( a j ) + cr-(aj)] j=l Under the set of constraints : A(aj, aj§ 2 ~, if aj P aj+1 and d(aj, aj+O = c~,if ai I aj§ n ai-1 EEW~ :1, w(j_>O, i E[n], j e[rr l], i=1 j cr § ( a j ) , t r - (a j) > O, j e[k] and ~ a small positive number The error function F expresses the over-estimations or under-estimations of the alternative actions over the weak-order. Its minimum value is the criterion for the control of the consistency degree between additive model and initial preference structure. The optimum value is F*=0. The consistency degree is also controlled by the index r of Kendall, where its value in range [-1, 1] expresses the distance between the two above rankings. 292 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 4.4 Stock Ranking System Methodology The stock ranking methodology is described in the following diagram in "Figure 2". Figure 2 : Stock Ranking Methodology . Initial Rankings. Three initial stock rankings are taking place. Each ranking is supported by the respective system shown in "Figure 3" : 9 Fundamental Analysis System. The system uses criteria which are indices from Fundamental Analysis as well as Qualitative indices. The resulting ranking provides the "stock position". 9 Technical Analysis System. The system uses indices from Technical Analysis as criteria. The resulting ranking provides the "stock trend". 9 Stock-Exchange Analysis System. The system uses stock-exchange indices as criteria. The resulting ranking provides stock-exchange "profile" of the stocks. G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 293 Figure 3: Initial Rankings . Lowest Stock Profile (LSP) Selection for every ranking. The lowest stock profile describes the lowest standards (least desirable criteria values) that a stock should meet in order to be considered worth investing and to be included in the portfolio. The stocks which are ranking below the L S P are "cut off". There is a different L S P for every ranking. So there is : 9 LSP for the Fundamental Analysis ranking, 9 L S P for the Technical Analysis ranking, 9 L S P for the Stock-exchange Analysis ranking. The lowest stock profile is calculated as follows : For every criterion there is a value which is equal to the least desirable criterion value "Figure 4". Figure 4: Lowest Stock Profile fi The L S P derives from an additive utility function: U ~ = ~ P i U i i=l 294 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 where p; is the weight of criterion g;. The LSP can be defined : 9 either by the administrator of the system, 9 or by the user-investor. The system administrator can provide the system with a default selection LSP, according to his judgement. Thus, the user-investor will be able to choose between : strict, regular and lenient LSP. On the other hand, the userinvestor will be able to determine his own specific LSP. . Stock Cutting-Off. All the stocks in every ranking that are below the LSP in the perspective ranking will be cut off from the stock list as "rejected", as shown in "Figure 5". Cutting off will take place at the point where : 9 for the Fundamental Analysis ranking : U/i) > UF >= Ul(i+l) 9 for the Technical Analysis ranking : U2O) > U2~ >= U2~+1) 9 for the Stock-exchange Analysis ranking : U3(k) > UJ' >= U3(k+l) Apparently, the remaining stocks in each ranking list may be the same, or they may be different ones, since every ranking utilises different criteria. Figure 5: Stock Cutting Off 4. Stocks Set Selection. At this point the user-investor has a number of altemative possibilities" 9 To select all the remaining stocks which appear in the F.A. (El), T.A. (E2) and S-E.A. (E3) rankings. In this case the set of stocks is 9E = E1 ~3 E2 • E3. In this set there are stocks with: utility F.A. = 0. These are stocks that were cut off from the F.A. ranking as rejected, but are remaining in the T.A. or S.A. ranking. G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 295 utility T.A. = 0. These are stocks that were cut off from the T.A. ranking as rejected, but are remaining in the F.A. or S.A. ranking. They are stocks with good fundamental figures and good stockexchange profile but they are found in an unfavourable conjuncture for investment. utility S.A. = 0. These are stocks that were cut off from the S.A. ranking as rejected, but are remaining in the F.A. or T.A. ranking. They are stocks with good fundamental figures and in a favourable conjuncture ~)),rvpla, in terms of technical analysis, but they have some non-sufficient stock-exchange indices (e.g. P/E). To select only the common stocks that are found in the remaining parts of the F.A. (El), T.A. (E2) and S-E.A. (E3) rankings. In this case the set of stocks is : E -- E1 r3 E 2 ~ E3. It is a clearly smaller set of stocks than the previous one. To select an intermediate set of stocks, as regards the two sets mentioned previously. Some combinations for the intermediate set are : E = E l W ( E 2 ~ E 3 ) , E = ( E I u E 2 ) n E 3 , E = E l n (E2kg E3), etc. 5. Time Function Selection. A time function T=f(t) is selected from a set of time functions. This time function will represent, as accurately as possible, the impact of the investment horizon on the weighting of the above ranking lists. The investment horizon is determined by the investor's profile. 6. F.A., T.A. and S-E.A Ranking Weighting. Value r of the function T=f(t) that derives from the investor's investment horizon, is multiplied by the utilities of the first ranking [Ul(i)], while the utilities of the second ranking [U2(j)] are multiplied by (I-r). Thus, the utilities U'l(i), for the Fundamental Analysis, and U'20"), for the Technical Analysis, derive. The longer-term the investment horizon is, the higher the weight of the Fundamental Analysis stock list. On the contrary, the shorter-term the investment horizon is, the higher the weight of the Technical Analysis list (T.A.). As regards the utilities of the Stock-Exchange Analysis ranking, they are multiplied by the mean of the two previous values, which is always 0,5 [(~+1-'0/2]. In "Figure 6" the set E of the stocks that emerge, is the one selected in step 4 and the number of stocks it includes, may be m. The stock utility list for every ranking contains the same stocks (m). The stock appearance range is not based on descending utility for every ranking, but according to an ascending stock number. 296 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 Figure 6: Ranking Weighting 7. Undertaken Risk definition. Co-efficient beta derives from the investor's profile and represents part of the risk the investor is willing to undertake. Beta can be expressed in three forms : 1"~tform. b~<b<b~ : where b~ is the lowest limit and b~ is the highest limit of undertaken beta. The middle of the space is calculated as b, = (b~- b~)/2, so, undertaken beta is b~ = b~ + b~,.Based on undertaken beta (b~) and beta (b) o f stock 1, we can calculate the absolute value of the difference : Abi =/b~-bJ In this case, beta (AbO participates as a criterion in the global ranking. The scale for Abi of every stock, is continuous. The most desirable value is the lowest value of Ab~ and the least desirable value is the highest value ofAb~ 2 nd form. b< b~ 0 b> b~ : In this case beta does not participate as a criterion in the global ranking. However, it is used after the ranking for the cutting off of the respective stocks. 3 "a form. The investor does not express any preference regarding beta in the stage of stock ranking but may possibly do so in the selection and portfolio composition stage. 8. Global Ranking. This is the compromise of the above mentioned weighted rankings. F.A. (U' l(i)), T.A. (U'2(i)) ~r S-E.A. (U'3(i)) rankings are utilised now as criteria (criterion 1 for F.A., criterion 2 for T.A. and criterion 3 for S.A.). The value scale is continuous in all three criteria. The most desirable value is the highest utility value, and the least desirable is the lowest utility value. The fourth criterion is ,~b;, which is defined in step 7, and represents the undertaken risk in relation with the stock risk. The global stock ranking is carried out on the basis of these four criteria, as shown in "Figure 7". G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 297 Figure 7: GlobalRanking 4. 5 Stock Ranking Multicriteria Systems The proposed system comprises four ranking systems, as shown in "Figure 1". They are the following : 4.5.1 Fundamental Analysis Multicriteria System [FAMS] FAMS is a stock ranking system based on Fundamental Analysis. FAMS is based on data drawn on the Athens Stock Exchange (ASE), as well as the Association of Greek Institutional Investors (A GII). Fundamental Analysis follows a different evaluation structure for every different type of Accounting Plan. There are four evaluation structures corresponding to four types of Accounting Plan, and one evaluation structure for qualitative criteria. A data base of the FA ratios, for every Accounting Plan category is created based on this information. The output of the FA Company Ranking, is a ranking list, for every AP category "Figure 8". The FA Company Ranking will be updated on a quarterly basis. 298 OperationalResearch.An InternationalJournal/ Vol.3,No.3 / September- December2003 Figure 8: Company Ranking System based on Fundamental Analysis 4.5.2 Technical Analysis Multicriteria System [TAMS] TAMS is a stock ranking system based on Technical Analysis. A w ~ [ axot;~i.~ ~ 6 ~cTIvTechnical Analysis Indices data base. The output o f the TA Stock Ranking, is a ranking list "Figure 9". The T A Stock Ranking will be updated on a daily basis. Figure 9 : Stock Ranking System based on Technical Analysis G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 299 4.5.3 Stock-Exchange Analysis Multicriteria System [SAMS] SAMS is a stock ranking system based on Stock-Exchange Figures. It draws data from the Stock-Exchange Analysis Indices data base. The output of the SEA Stock Ranking, is a ranking list shown in "Figure 10". The SE Stock Ranking will be updated on a daily basis. Figure 10 : Stock Ranking System basedon Stock-ExchangeFigures 4. 5.4 Global Stock Ranking Multicriteria System [SRMS] The system shown in "Figure 11" is responsible for the global stock ranking. The inputs it receives are the rankings of the three above mentioned subsystems : FAMS, TAMS, SAMS, and in combination with undertaken risk as well as investment horizon, it carries out the global stock ranking. The Global Stock Ranking will be updated on a daily basis since two of the subsystems contributing in the evaluation (TAMS and SAMS), will be updated every day. 300 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 Figure 11 : Global Stock Ranking Multicriteria System 5. Concluding Comments The proposed system, by incorporating a huge volume of financial, stock-exchange and other relevant information, is a "live" system that operates in "real world" conditions, since its data are updated on a daily basis through multiple sources. The final system output is a global evaluation of the ASE stocks, in the form of a ranking list, from the best towards the worst stock. The system is intended to support investment decisions. It addresses both institutional and private investors with either a long-term or a short-term investment horizon. The special features - advantages of the system are outline as follows : 9 The proposed system is an Multicriteria DSS since it uses multicriteria analysis methodologies. 9 It comprises, as a whole, all three components which are necessary for a successful investment decision : - Market Psychology Fundamental Analysis - Technical Analysis - G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 301 It deals with risk (beta) in a unique way. Beta is no longer a basic criterion in the Fundamental Analysis. It is introduced, however, in the investor's profile, both at Stock Ranking level and at Portfolio Selection and Composition level. The proposed system can operate either autonomously or as part of an integrated system that incorporates all portfolio management operations. It presents a rather high automation level, as well as a high level of participation and interaction as far as the decision-maker is concerned, within the decisionmaking process. References Achelis, St. (2000), Technical Analysis from A to Z, McGraw-Hill. Colson, G. and De Bruyn C. (1989), "Models and Methods in Multiple Objectives Decision Making", Mathematical Computing and Modelling, Vol. 12(11/12), 1201-1211. Conso, P. (1981), la gestion financiere de l'entreprise : les techniques et l'analyse financiere (tome 1), Dunod. Depallens, (3. (1980), gestionfinanciere de l 'entreprise, Editions Sirey. Devaud, J.M., Groussaud, (3. and Jacquet-Lagr~ze, E. (1980), "UTADIS: Line m6thode de construction de fonctions d' utilit~ additives rendant compte de jugements globaux", European Working Group on Multicriteria Decision Aid, Bochum. Edwards, R.D. and J. Magge (1992), Technical Analysis of Stock Trends, John Magee Inc, Boston. Giese, F. (1981), Pratique De L'Analyse Financiere Des Bilans, CLET, Editions BANQUE. Gualino, Ph. and M. Remilleret (1979), Entrainement a l'Analyse des Bilans, CLET, Editions BANQUE. Hurson, C. and Zopounidis, C. (1997), "Gestion de Portfeuilles et Analyse Multicrit~re", Economica, Paris. Jacquet-Lagr6ze, E. and Siskos, J. (1982), "Assessing a set of additive utility functions for multicriteria decision-making: The UTA method", EJOR, Vol. 10, 151164. Jacquet-Lagr6ze, E. and Siskos, J. (2001), "Preference disaggregation: 20 years of MCDA experience", European Journal of Operational Research, Vol. 130, 233245 302 Operational Research. An International Journal / Vol.3, No.3 / September- December 2003 Keeney, R. and Raiffa, H. (1976), "Decision with Multiple Objectives:Preferences and Value Tradeoffs", John Wiley & Sons. Keeney, R. L. (1992), Value Focused Thinking, Harvard University Press, Cambridge, MA. Kostkowski, M. and Slowinski, R. (1996), "UTA+ Application (V. 1.20)-User's Manual. Document du LAMSADE, No. 95, Universit6 de Paris-Dauphine, Paris. Lee, S.M. and Chesser, D.L. (1980), "Goal programming for portfolio selection", The Journal of Portfolio Management, Spring, 22-26. Markowitz H. (1952), "Portfolio selection", Journal of Finance, 7 (1), 77-91. Murphy, J.(1986), Technical Analysis of the Financial Markets, NY: New York Institute of Fianance. Nison, St. (1991), Japanese Candlstick Charting Techniques, NY: New York Institute of Fianance. Pardalos, P., M. Sandstrom and C. Zopounidis (1994), "On the use of optimization models for portfolio selection: a review and some computational results", Computational Economics, vol. 7, no. 4, 227-244. Roy, B. (1968), "Classement et choix en prrsence de points de vue multiples (la mrthode Electre)", Revue Francaise d'Informatique et de, Vol. 8, 57-75. Roy, B. (1978), "ELECTRE III : algorithme de classement bas6 sur une reprrsentation flou des prrfrrences en prrsence de critrres multiples", Cahiers du CERO, Vol. 20(1), 3-24. Roy, B. and Bertier, P. (1971), "La mrthode ELECTRE II", Note de travail, No. 142, SEMA. Samaras, G.D. and N.F. Matsatsinis (2003), "INTELLIGENT INVESTOR : An Intelligent DSS for Portfolio Management ", 16th National Conference on Operational Research in Project Management, TEI of Larissa, September 25-27, 2003. Sharpe, W.F (1964), "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk", The Journal of Finance, Vol. 19, No. 3, pp. 425-442. Siskos, J. and Despotis, D. K. (1989), "A DSS oriented method for multiobjective linear programming problems", Decision Support Systems, Vol. 5, 47-55. Siskos, J. and Matsatsinis, N.F. (1993), "A DSS for market analysis and new product design", Journal of Decision Systems, Vol. 2(1), 35-60. Siskos, Y. (1980), "Comment modrliser les prrfrrences au moyen de fonctions d' utilit6 additives ~, RAIRO Rrcherche Oprrationnelle, Vol. 14, 53-82. G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation 303 Siskos, Y. and Yannacopoulos, D. (1985), "UTASTAR: An ordinal regression method for building additive value functions", Investigaq~o Operacional, Vo[. 5(1), 39-53. Siskos, Y., Grigoroudis, E., Zopounidis, C. and Saurais, O. (1998), "Measuring customer satisfaction using a syrvey based preference diasaggregation model", Journal of Global Optimization Vol. 12(2), 175-195. Siskos, Y., Spyridakos, A. and Yannacopoulos, D. (1993), "M1NORA: A Multicriteria Decision Aiding System for Discrete Alternatives", Journal of Information Science and Technology, in Siskos, Y and C. Zopounidis (eds), Special Issue on Multicriteria Decision Support Systems, Vol. 2(2), 136-149. Siskos, Y., Spyridakos, A. and Yannacopoulos, D. (1999), "Using artificial intelligent and visual techniques into preference disaggregation analysis: The MIDAS system", European Journal of Operational Research, Vol. 113,236-246. Spronk, J. and W. Hallerbach (1997), "Financial modeling: Where to go? With an illustrator for portfolio management", EJOR, No 99, pp. 113-125. Turban, E. and Aronson, E. J. (2001), "Decision Support Systems and Intelligent Systems", Prentice Hall International, Inc. Vincke, Ph. (1992), '~MulticriteriaDecision Aid", Wiley, New York. Vizzanova, P. (1981), Gestion Financiere : Analyse Statique, Analyse Dynamique, Librairies Techniques. Vranes, S., Stanojevic, M., Stevanovic, V. and Lucin, M. (1996), "INVEX: Investment advisor expert system", Expert Systems Vol. 13(2), 105-119. Weston, F. and E. Brigham (1986), Baatx~q Apx~ rrl~Xprl#aToolxovoplr/l~ A la;(~iplar/r ~al Floktzt~lg, Elc~boeE 1-Iana~lOrl. Yu, W. (1992), "ELECTRE TRI: Aspect m6thodologiques et manuel d' utilization", Document du Lamsade, 74, Universit6 Paris-Dauphine. Zopounidis C., Matsatsinis, N. F. and Doumpos, M. (1996), "Developing a multicriteria knowledge-based decision support system for the assessment of corporate performance and viability: The FINEVA system", Fuzzy Economic Review 1/2, 3553. Zopounidis, C. and Doumpos, M. (1998), "Developing a multicriteria decision support system for financial classification problems: the FINCLAS system", Optimization Methods and Software, Vol. 8, 277-304. Zopounidis, C. and Doumpos, M. (1999), "Business failure prediction using UTADIS multicriteria analysis", Journal of the Operational Research Society, Vol. 50(11), 1138-1148. 304 Operational Research. An International Journal / Vol.3, No.3 / September - December 2003 Zopounidis, C. and Doumpos, M. (2000a), "PREFDIS: A multicriteria decision support system for sorting decision problems", Computers and Operations Research, Vol. 27(7-8), 779-797. Zopounidis, C. and Doumpos, M. (2000b), "INVESTOR: A decision support system based on multiple criteria for portfolio selection and composition", in A. Colomi, M. Paruccini and B. Roy (eds.), A-MCDA-A, European Commission, Joint Research Centre, 371-381. Zopounidis, C., Despotis, D.K. and Kamaratou, I. (1998), "Portfolio Selection Using the ADELAIS Multiobjective Linear Programming System", in Computational Economics (Kluwer Academic Press), Vol. 11,189-204. Appendix I : Fundamental Analysis Structure a Commerce/lndustryFundamentalAnalysis 9 Study of the Financial Structure of an Enterprise o Assets & Liabilities Structure o Financial Balance o Liquidity o Debt Ratio 9 Management Performance 9 Profitability 9 Self-financing Policy 9 Qualitative Criteria 121 Banks Fundamental Analysis 9 Capital Adequacy & Liquidity o Liquidity o Capital Adequacy 9 Investment Policy & Deposit Structure o Investment Policy o Deposit Structure 9 Profitability & Income o Profitability o Net Income 9 Qualitative Criteria Q Insurance Companies Fundamental Analysis 9 Profitability o Capital Profitability G.D. Samaras, N.F. Matsatsinis, C. Zopounidis / A multicriteria DSS for a global stock evaluation o Profit Margin 9 Qualitative Criteria [:1 Investment Companies Fundamental Analysis 9 Investment Company Performance o Net Asset Value Performance o Premium/Discount o Net Asset Value Change 9 Qualitative Criteria I:1 Qualitative Criteria 9 9 9 9 9 9 Managers' Experience Position of the firm in the market Technological Structure of the firm Organization-Personnel Special Competitive Advantages of the firm Flexibility of the firm towards the tendency of the market Appendix 2 : Technical Analysis Indices 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 Low / High Price Opening / Closing Price Trading Volume Trading Value SMA : Simple Moving Average 150-days SMA / 30-Days SMA / 9-days SMA EMA : Exponential Moving Average 30-days E M A 20-days High / 4-days High 20-days Low/10-days Low Previous Top / Bottom OBV : On Balance Volume 14-days Simple Moving Average (OBV) R S I : Relative Strength Index RSI-14 : 14-daysRSI 20-days SMA(RSI-14) MACD : Moving Average Convergence/Divergence 9-days SMA(MACD) MAX(Prices) / MIN(Prices) 305 306 Operational Research. An International Journal / Vol.3, No.3 / September- December 2003 Appendix : Stock-Exchange Analysis Indices Q Q EPS : Earning Per Share PER : Price Earning Ratio DPS : Dividend Per Share ROE 9Return On Equity P/BV : Price per Book Value Marketability Beta Coefficient Shares Number