Application 3 – Annotated Bibliography

Multicriteria framework for selecting a process modelling language

Ana Carolina Scanavachi Moreira Campos and Adiel Teixeira de Almeida*

Department of Management Engineering, Federal University of Pernambuco, Recife, Brazil

(Received 24 October 2012; accepted 7 March 2014)

The choice of process modelling language can affect business process management (BPM) since each modelling language shows different features of a given process and may limit the ways in which a process can be described and analysed. However, choosing the appropriate modelling language for process modelling has become a difficult task because of the availability of a large number modelling languages and also due to the lack of guidelines on evaluating, and comparing languages so as to assist in selecting the most appropriate one. This paper proposes a framework for selecting a modelling language in accordance with the purposes of modelling. This framework is based on the semiotic quality framework (SEQUAL) for evaluating process modelling languages and a multicriteria decision aid (MCDA) approach in order to select the most appropriate language for BPM. This study does not attempt to set out new forms of assessment and evaluation criteria, but does attempt to demon- strate how two existing approaches can be combined so as to solve the problem of selection of modelling language. The framework is described in this paper and then demonstrated by means of an example. Finally, the advantages and disadvantages of using SEQUAL and MCDA in an integrated manner are discussed.

Keywords: business process management; process modelling language; semiotic quality framework; multicriteria decision aid; ELECTRE TRI-B

1. Introduction

In recent years, businesses have been facing the challenges of globalisation, competitive- ness and rapid changes in business environments and market requirements. As a result, organisations are paying more attention to supporting business process management (BPM) as it is able to adapt quickly to these new business environments and requirements and also so that they can maintain their competitive advantage and achieve their perfor- mance goals.

BPM can be defined as a management approach that uses methods and tools to support the design, analysis and control of business processes. The aims are to enhance customer satisfaction and product quality as well as to improve operations (Elzinga et al. 1995). BPM has become an important topic in the competitiveness of enterprise informa- tion systems, with many new patents issued (Campos, Daher, and de Almeida 2011).

The BPM life cycle consists of the following phases (Weske 2007): design and analysis, configuration, enactment and evaluation. All phases are important. However, the design phase, the stage during which the process is modelled, deserves special attention since serious problems may arise in other BPM phases, if the design phase is not conducted correctly.

*Corresponding author. Email: [email protected]

Enterprise Information Systems, 2016 Vol. 10, No. 1, 17–32, http://dx.doi.org/10.1080/17517575.2014.906047

© 2014 Taylor & Francis

Business process modelling is a tool which aids the description of business pro- cesses. This modelling is necessary in order to understand, evaluate, analyse and improve business processes (Tan et al. 2008; Vergidis, Turner, and Tiwari 2008; Xu 2011; Xu et al. 2012; Viriyasitavat, Xu, and Martin 2012). Nowadays, there are a large number of modelling languages that have been used to describe business processes, such as business process modelling notation (BPMN), unified modelling language (UML) activity-diagrams, Petri Net, event-driven process chains (EPC), YAWL (Xu et al. 2008, 2009; Figl, Mendling, and Strembeck 2009; Capozucca and Guelfi 2010; Quartel, Steen, and Lankhorst 2012). Process modelling is not related to data modelling in information systems. The process regards the organisation in which the information has to be provided. The data modelling is a step to be considered in another study.

As each language shows different features of a given process (Vergidis, Turner, and Tiwari 2008; Xu 2011; Li et al. 2013; Tan et al. 2013) and may limit the ways in which a process can be described and analysed, the choice of process modelling language can affect BPM (Luo and Tung 1999). Thus, choosing the appropriate language for modelling has become a complex task not only because of the availability of a large number of modelling languages but also due to the lack of guidelines on evaluating and comparing such languages in order to assist in selecting the most appropriate ones.

In this paper, a framework is put forward to help the manager/modeller (decision- maker (DM)) to evaluate and select a modelling language for BPM. This framework is based on semiotic quality framework (SEQUAL) for evaluating modelling languages and on the ELECTRE TRI-B multicriteria method so as to classify process modelling lan- guages as per the purposes of the modelling. This classification puts languages with similar features into the same category which can give the manager/modeller more than one choice of modelling language that meets his/her modelling purposes. After classifying the languages, the manager/modeller can deepen his/her knowledge in those that most meet his/her modelling purposes and then make a more adequate choice.

This introductory section presents the context in which the paper is set as well as the aim of paper. Section 2 gives a short description of the SEQUAL. Section 3 discusses related work. Section 4 describes some concepts of multicriteria decision aid (MCDA) and the ELECTRE TRI-B method used in the framework put forward in this paper. Section 5 describes this framework and how it will help the manager/modeller to select a modelling language for BPM. Section 6 provides a demonstration of the framework by means of an illustrative example. In Section 7, there is some discussion of the framework, and finally Section 8 draws some conclusions and makes suggestions for future research.

2. Evaluating languages using SEQUAL

In this section, we give a short description of the SEQUAL for evaluating languages. The SEQUAL is used for evaluating quality of models and then was extended so it

could be used to assess the quality of modelling languages (Aagesen and Krogstie 2010). According to La Rosa et al. (2011), the SEQUAL is one of three major frameworks for evaluating process modelling languages. The other two are the workflow patterns frame- work (van der Aalst et al. 2003) and the Bunge–Wand–Weber (BWW) framework (Wand and Weber 1993).

A review of the literature shows that the quality framework has been used to evaluate and compare some modelling languages. Krogstie (2003) uses the quality framework to evaluate UML. An evaluation of BPMN using this framework was conducted by Wahl and Sindre (2005). Cortes-Cornax et al. (2011) extended the quality framework to

18 A.C. Scanavachi Moreira Campos and A.T. de Almeida

evaluating choreographies in BPMN 2.0. Both Nysetvold and Krogstie (2005) and Krogstie and Arnesen (2005) applied the quality framework to evaluate and compare modelling languages. Nysetvold and Krogstie (2005), for example, evaluated and com- pared UML, BPMN and extended enterprise modelling language (EEML).

The quality framework with a focus on language quality evaluates the appropriateness of the language for process modelling in six quality areas: organisational appropriateness, domain appropriateness, tool appropriateness, participant appropriateness, modeller appropriateness and comprehensibility appropriateness (Nysetvold and Krogstie 2005). These areas result from relationships among the following sets (Figure 1) (Nysetvold and Krogstie 2005):

● Modelling goal – the goals of the modelling; ● Language extension – what can be expressed in the modelling language based on

the graphemes, vocabulary and syntax of the language used; ● Domain – the set of statements which can be expressed about the real situation; ● Externalising the model – the set of all statements expressed in the model by the

language;

Language extension

Modelling goal

Domain

Technical actors’ interpretation

Stakeholders’ explicit knowledge

Modellers’explicit knowledge

Model externalisation*

Organisational appropriateness

Domain appropriateness

Tool appropriateness

Participant appropriateness

Modeller appropriateness

Comprehensibility appropriateness Social actors’ interpretation

Figure 1. Language quality based on quality framework (Adapted from Nysetvold and Krogstie (2005)). * In the language quality framework, the language extension set is not related to the model externalisation set.

Enterprise Information Systems 19

● Stakeholders’ explicit knowledge – the explicit knowledge of the stakeholders involved in the modelling;

● Modellers’ explicit knowledge – the explicit knowledge of the modellers involved in the modelling;

● Social actors’ interpretation – the set of statements that the participants interpret from the model;

● Technical actors’ interpretation – the interpretation of the statements in the model by different modelling languages.

The organisational appropriateness quality area is related to the needs of the organisation in the modelling context. Domain appropriateness measures the capacity of language to express everything in the domain. Tool appropriateness is related to the technical inter- pretation; if the language allows automatic reasoning. Thus, different aspects analysed in tool appropriateness are a mean to achieve syntactic, semantic and pragmatic quality. Participant appropriateness measures stakeholders’ knowledge about the language and their ability to use it. Modeller appropriateness measures the capacity of language to express statements about all the knowledge that the modeller has. Comprehensibility appropriateness is related to how well the participants interpret and understand the language (Wahl and Sindre 2005). In this area is possible to evaluate, for example, if the symbol discrimination is easy, if the concepts of the language is easily distinguishable from each other, if the number of concepts is reasonable. See Nysetvold and Krogstie (2005) for a more detailed description of the quality framework.

3. Previous related work

Although there are many studies in the literature that propose ways to evaluate and compare modelling languages, there are few which actually propose a way of choosing the modelling language. Even among those which propose a way, none presents a consistent method for selecting the modelling language.

Luo and Tung (1999), for example, designed a framework for evaluating modelling languages and selecting one from all those evaluated. By following this framework, the language selected should be the one that has the greatest number of required character- istics, set by the modelling objectives. However, no decision-making technique is estab- lished. In the example given in their paper, language selection was made by comparing the table of the importance of the characteristics required of the language needed with the table of the characteristics of each language evaluated.

Krogstie and Arnesen (2005) used the quality framework to evaluating five modelling languages. Based on this evaluation, they select a modelling language to be used by an oil company. The selection process is given simply by the sum of the scores received by each language. The languages with the highest sum are considered the most appropriate for modelling. Similarly, Nysetvold and Krogstie (2005) applied the quality framework in an insurance company. In their paper, three modelling languages were compared.

On the basis of the fact that in a selection process, it is essential to know and understand the strengths and weaknesses of each alternative as well as to carry out the comparison of such alternatives, many researchers have only provided evaluating and comparing modelling languages with a view to helping select a language.

These include Söderström et al. (2002) who developed a framework for supporting the comparison of modelling languages and making this task easier. According to them, the resulting comparison matrix of the framework aids in selecting a modelling language since

20 A.C. Scanavachi Moreira Campos and A.T. de Almeida

it gives a comprehensive view of modelling languages and point out differences among them. In order to demonstrate the framework, three modelling languages were compared.

After choosing a modelling language without performing a detailed investigation, Glassey (2008) decided to compare the three modelling languages that were previously suggested so as to conduct process modelling. His/her objective with this research was to provide a framework for supporting detailed analysis before making decisions.

Recker et al. (2009) consolidate a comprehensive view of the representational cap- abilities of seven modelling languages in a table and suggest that it can be used to support selecting a modelling language.

Aguilar-Savén (2004) proposed a classification of process modelling techniques according to their purpose (learning, support to design, support to conducting a process, IT support) with the aim of helping practitioners and academics to choose the most appropriate technique. Giaglis (2001) developed a taxonomy for helping users evaluate and choose a modelling technique in line with the requirements of a project.

In the context of computer and information science, Su and Ilebrekke (2002) used the semiotic quality framework to evaluate and compare ontology languages and tools in order to aid selecting a proper language and tool for building anthologies in a specific domain.

Liu et al. (2008) provided an overview of several modelling languages in the context of enterprise application integration (EAI) and conducted an analysis of enter- prise application considering both semantic and syntactic integration. Finally, some issues which need to be addressed in the future are identified. Among these issues is the selection of an enterprise application including a process modelling system.

In the area of outsourcing, Chu et al. (2002) drew up an approach for external partner selection. They used the analytic hierarchy process multicriteria method for selecting the best partner from the potential ones.

4. MCDA and choice of the multicriteria method

In this section, an overview of MCDA and the ELECTRE TRI-B method is given as is the justification for choosing the ELECTRE TRI-B multicriteria method.

MCDA is an approach that compares the alternatives that are being analysed in the decision process according to a variety of criteria with the aim of stating preference relationships among these alternatives. The criteria represent the attributes or points of view that are used to evaluate the set of alternatives (Figueira, Greco, and Ehrgott 2005; Perçin 2010).

In order to deal with multiple criteria decision problems, some methods have been developed which can be classified into methods that add criteria in a unique function of synthesis, such as multiattribute utility theory and outranking methods, including ELECTRE and PROMETHEE, and interactive methods (Belton and Stewart 2002).

These multicriteria methods have been applied in a variety of problems related to different topics, enhancing contribution between MCDA with other areas of knowledge. This has added to scientific knowledge, since these integrations require adaptations and adequate interconnection among areas. This includes a diversity of subjects, such as contracts and outsourcing (de Almeida 2007; Brito, de Almeida-Filho, and de Almeida 2010), risk management (Brito, de Almeida, and Mota 2010), project management (Mota, de Almeida, and Alencar 2009; Mota and de Almeida 2012) and water management (Silva, Morais, and de Almeida 2010; Morais and de Almeida 2006).

Methods that add criteria in a unique function of synthesis are compensatory methods in which it is possible to offset a disadvantage of one criterion by an

Enterprise Information Systems 21

advantage it has over other criteria. Such methods favor alternatives that have a low performance in some criteria but a high one in others. This type of compensation does not occur with outranking methods, so they are not compensatory (de Almeida 2007). Interactive methods consist of phases, and after each calculation phase, a solution (or several solutions) is (are) presented to the DM. The DM affirms if this solution is the end result or he/she provides more information to run a new calculation phase (Vincke 1992; Belton and Stewart 2002).

When choosing the multicriteria method most appropriate for a problem, account needs to be taken of what context the problem is inserted in, its characteristics, the DM’s structure of preference as well as what type of outcome is intended for the problem since each method gives a different type of result. Roy (1996) specifies four basic problematic types that correspond to four types of outcomes: choice, sorting, ranking and description.

In order to select the appropriate modelling language for BPM, we consider that an outranking method is more appropriate because the alternative(s) indicated by this type of approach will have better average performance(s) instead of having an excellent perfor- mance in some criteria, but a poor one in others. Furthermore, the characteristics of this problem make it more suitable as a sorting problem since sorting multicriteria methods assign alternatives to different categories only if they are really different. This feature is useful since the data and estimated parameter values are imprecise and are based on qualitative evaluations. In addition, as the assignment of an alternative to a category by the ELECTRE TRI-B method depends only on its own value and not on comparing the alternative to other alternatives from the set (as occurs in ELECTRE TRI-C), this method is used in the model set out in this paper and is outlined below (Figueira, Greco, and Ehrgott 2005; Mota, de Almeida, and Alencar 2009, Mota and de Almeida 2012; Brito, de Almeida, and Mota 2010).

4.1. Electre TRI-B method

The ELECTRE TRI-B is an outranking method developed to deal with sorting problems. This method assigns alternatives (a1,…,ai,…,as) to predefined ordered categories (C1,…, Ck, Ck + 1,…, Ck + m) by using the results of comparing each alternative with the reference profiles (Ir1,…, Irk, Irk + 1,…, Irk + (m−1)), which define the limits of the categories (Figure 2) (Figueira, Greco, and Ehrgott 2005; Mota and de Almeida 2012).

Category C k+3

Category C k+2

Category C k+1

Ir k

Ir k+2

Ir k+1

Category C k

Figure 2. Categories and their limits (Adapted from Merad et al. (2004)).

22 A.C. Scanavachi Moreira Campos and A.T. de Almeida

To do so, the ELECTRE TRI-B method constructs outranking relations between each alternative and reference profile and then explores them.

Let gj be a criterion (j = 1,…n, where n is the number of criteria), the outranking relation S is built from paired comparisons between each alternative and reference profile (gj(ai), gj(Irk) and (gj(Irk), gj(ai)) for all criteria and is established by calculating the partial concordance indices cj, the global concordance index C, the discordance index dj and finally the credibility index σ(ai,Irk). For more details, see Merad et al. (2004). The credibility index σ(ai,Irk) expresses the intensity with which it is possible to believe that ai is at least as good as the profile Irk.

In order to reduce uncertainties in evaluations (gj(ai) and gj(Irk)), the indifference qj and the preference threshold pj can be taken into account. qj(Irk) defines the largest value of difference between evaluations in a criterion gj that ensures indifference between ai and Irk and pj(Irk) indicates the smallest difference between these evaluations consistent with a preference for ai (Merad et al. 2004).

In constructing the outranking relation, the DM needs to define a set of weights (w1, w2,…, wn) – one for each criterion – that expresses the relative importance among criteria and is used in the concordance test, and a set of veto thresholds (v1(Irk), v2(Irk),…, vn(Irk)) that is employed in the discordance test. The veto threshold expresses the smallest difference between the evaluation of the reference profile and alternative (gj(Irk) – gj (ai)) that negates the assertion ‘ai outranks Irk’ (Lourenço and Costa 2004).

The outranking relation is established by analysing the value of credibility index (σ) with the cutting level (λ). The cutting level is a value situated in the interval between 0.5 and 1 and means the smallest value that the credibility index can be in order to affirm that ‘ai outranks Irk’ (aiSIrk) or ‘Irk outranks ai’ (IrkSai) and is defined by the DM. The alternatives ai and reference profiles Irk may relate to each other as follows: (Mousseau and Slowinski 1998)

(1) σ(ai, Irk) ≥ λ and σ(Irk, ai) ≥ λ: ai is indifferent to Irk; (2) σ(ai, Irk) ≥ λ and σ(Irk, ai) < λ: ai outranks Irk; (3) σ(Irk, ai) ≥ λ and σ(ai, Irk) < λ: Irk outranks ai; (4) σ(ai, Irk) < λ and σ(Irk, ai) < λ: ai is incomparable to Irk.

Outranking relations are explored in order to allocate the alternatives to one of the categories. There are two classification procedures which are called pessimistic and optimistic, where the difference between them is in the sequence of the comparison of the alternative and reference profiles (Zopounidis and Doumpos 2002).

The pessimistic procedure is more demanding and assigns each alternative to the highest category for which the alternative outranks the reference profile of the lower category. The optimistic process is less demanding, and assigns each alternative to the lowest category for which the higher category reference profile outranks the alternative (Lourenço and Costa 2004).

Divergence between two classifications can happen and arises from incomparability situations between an alternative and a reference profile. Thus, in such a situation, the DM can adopt one of two classifications (Mousseau and Slowinski 1998). When the DM wishes to be prudent, the pessimistic procedure should be adopted.

5. Framework for selecting a process modelling language for BPM

The framework proposed in this paper for evaluating and choosing a modelling language is based on the SEQUAL and MCDA, particularly the ELECTRE TRI-B multicriteria method.

Enterprise Information Systems 23

This framework aims to help the manager/modeller (DM) selects an appropriate modelling language for BPM since by using it, he/she can identify among all modelling languages analysed, those that best meet his/her modelling purposes. Thereafter, the manager/modeller can deepen his/her knowledge in such languages and then make a more adequate decision. The framework consists of six steps which are described below (Figure 3).

● First step

It is essential to know what you would get from modelling before selecting a modelling language. Consequently, the first step consists of determining the purpose of modelling, i.e. its objective. This purpose can be, for example, either to understand the structure and dynamics of the process or to capture the current process in order to identify opportunities for improvement or to ensure that all employees, suppliers and customers have a common understanding of the process or to control and monitor process activities and so on. The purpose does not necessarily correspond to a single objective; depending on the situation, there can be a combination of several objectives. Thus, for instance, the purpose may be both to understand the process structure and to implement a process control system.

● Second step

In the second step, the modelling languages to be evaluated should be identified.

● Third step

In this step, each modelling language should be evaluated by means of the SEQUAL. As described above, the SEQUAL evaluates the appropriateness of the language for model- ling in six quality areas.

Define the purpose(s) of modelling

Establish the list of modelling languages to be evaluated

Evaluate languages by means of the Semiotic Quality Framework

Apply multicriteria method

Analyse the results of classification and deepen understanding of the modelling languages assigned to the best category

Select the modelling language

Figure 3. Framework for selecting an appropriate modelling language.

24 A.C. Scanavachi Moreira Campos and A.T. de Almeida

However, the SEQUAL is too general and too abstract to allow proper evaluation of the modelling languages. So, it is necessary to include a list of criteria and requirements related to each of the six areas of language quality in order to evaluate the modelling languages. Nysetvold and Krogstie (2005) propose the use of a list of criteria defined by Østbø (2000) who identified about 60 potential criteria/requirements.

Thus, the manager/modeller can evaluate the modelling languages against those criteria/requirements that he/she believes are relevant according to the purpose of model- ling previously defined, so it is not necessary to evaluate the modelling languages against all 60 criteria/requirements. Moreover the manager/modeller can add other criteria/ requirements to this list.

In order to evaluate the modelling languages, a 4-point scale from 0 to 3 should be used as Nysetvold and Krogstie (2005) propose. According to this scale, any criterion/ requirement which is not supported by the language has grade 0. It has grade 1 if partly supported, grade 2 if satisfactory supported and 3 if is very well supported.

● Fourth step

This step consists of applying the multicriteria method so as to classify the modelling languages according to their appropriateness for conducting the modelling. As explained before, the ELECTRE TRI-B multicriteria method is the most appropriate one for this kind of problem and therefore is used in the framework proposed in this paper. A tool available on the LAMSADE site (www.lamsade.dauphine.fr/) or Microsoft Office Excel software can be used in order to obtain the classification of the modelling languages.

The ELECTRE TRI-B method requires that some data, such as criteria weights and the limit of the categories, and some parameters be defined. The criteria weights represent their relative importance and their value increases in line with the importance of the criteria. Thus, the manager/modeller should express the importance of each criterion/ requirement taking into account its contribution towards achieving his/her modelling purpose(s) (defined in Step 1). The criteria/requirements that are not used for evaluating the modelling languages automatically receive weight 0.

The categories that will serve as standards to classify the modelling languages should be defined too. As each category is limited by two reference profiles, an upper and a lower one, the values of these indices need to be specified. The categories will be defined as per the manager’s/modeller’s preferences.

Finally, the values of parameters need to be determined. They must be defined as per the manager’s/modeller’s preference structure. Thus, in order to reduce the imprecision of the evaluations of the modelling languages, the manager/modeller can establish the values for the preference (p) and indifference (q) thresholds for each criterion. He/she should also define the veto threshold (v) for each criterion. As mentioned above, that parameter is related to the idea of veto in relation to the assertion that an alternative outranks a reference profile. The cutting level (λ) has also to be determined. The cutting level should be defined carefully since it plays an important role in classifying the alternatives. If its value is high (close to 1), the alternatives will be allocated to the lower categories by the pessimistic procedure while the optimistic procedure will allocate them to the higher categories. If the cutting level value is low (close to 0.5), the inverse will happen.

Thereafter, the conduct of a sensitivity analysis is required to analyse how sensitive the result is by changing the values of the parameters (p, q, v, λ).

Enterprise Information Systems 25

● Fifth step

In this step, the result of the classification should be analysed. If there is more than one modelling language assigned to the best category, the manager/modeller needs to deepen his/her understanding of such modelling languages in order to have better knowledge of their benefits before selecting one. Nevertheless, it is worth pointing out that any one of those languages is in line with the manager’s/modeller’s modelling purpose(s).

● Sixth step

The manager/modeller selects the language which will be used in business process modelling.

6. Numerical application

In order to illustrate the framework proposed, we used some data and information about the case of an insurance company described in Nysetvold and Krogstie (2005), but we have considered another DM (manager), instead the previous one. Similarly, our numer- ical application has the same purposes of the modelling, and evaluates the same modelling languages as did Nysetvold and Krogstie (2005).

Thus, as in Nysetvold and Krogstie (2005), the purposes of modelling are related to implementing a system that integrates different business systems present in different departments of the company and supports employees in understanding the aspects of the company and in communicating this to other employees (Step 1). The process modelling languages evaluated are the EEML, UML 2.0 activity-diagrams and BPMN (step 2). Each modelling language was evaluated by means of the SEQUAL using a list of criteria defined by Østbø (2000) (Step 3). The table showing how the modelling languages were evaluated is presented in Nysetvold and Krogstie (2005).

After evaluating the modelling languages, the ELECTRE TRI multicriteria method is applied in order to classify the modelling languages according to their appropriateness for conducting the modelling (Step 4).

To define the weights, the new DM used a 6-point scale from 0 to 5 (0 being the least important and 5 being the most important). The importance of each criterion/ requirement was established taking into account its contribution towards achieving the modelling purpose(s). Table 1 presents the weights obtained from the manager’s judge- ments, bearing in mind that the criteria/requirements that are not used for evaluating the modelling languages automatically receive a weight of 0, and hence they were not inserted in Table 1.

In this case, the manager decided that the modelling languages should be classified into three categories according to their appropriateness for conducting the modelling:

Category 1: ‘Most Suitable’: the languages assigned in this category are appropriate for performing the modelling;

Category 2: ‘Possibly Suitable’: the languages assigned in this category may be appropriate for performing the modelling;

Category 3: ‘Not Suitable’: the languages assigned in this category are not appropriate for performing the modelling.

26 A.C. Scanavachi Moreira Campos and A.T. de Almeida

As there are three categories, two reference profiles (Ir1 and Ir2) should be established. Reference profile Ir1 determines the minimum performance a language must achieve to be in Category 1, and reference profile Ir2 defines the minimum performance a language must achieve to be in Category 2. Otherwise the language is allocated to Category 3. These reference profiles (Table 1) were determined based on the manager’s preference structure in accordance with the scale used to evaluate the modelling languages. Due to the fact that the better the language is evaluated by using the given criteria, the more appropriate and complete it is, and therefore reference profile (Ir1) which defines the first category – the ‘Most Suitable’ category – has a higher value than reference profile (Ir2).

With regard to the parameters:

● Due to the manager’s preference structure, preference (p) and indifference (q) thresholds and the veto threshold for each criterion were not applied.

● The cutting level (λ) was established by the manager at 0.6.

Table 1. Weights of the criteria and reference profiles.

Criterion/Requirement number Weight Ir1 Ir2

1.1 5 2 1 1.2 2 2 1 1.3 1 2 1 1.4 2 2 1 1.5 2 2 1 1.6 2 2 1 1.7 2 2 1 1.8 1 2 1 1.9 1 2 1 1.10 1 2 1 1.11 1 2 1 2.1 4 2 1 2.2 2 2 1 2.3 1 2 1 2.4 2 2 1 2.5 2 2 1 4.1 2 2 1 4.2 1 2 1 4.3 2 2 1 4.4 4 2 1 4.5 3 2 1 4.6 2 2 1 4.7 2 2 1 5.1 2 2 1 5.2 4 2 1 5.3 1 2 1 5.4 2 2 1 5.5 3 2 1 6.1 4 2 1 6.2 2 2 1 6.3 4 2 1 6.4 3 2 1

Enterprise Information Systems 27

Based on all this information, the ELECTRE TRI-B method was run on Microsoft Office Excel software. As the new DM is cautious, he/she adopted the pessimistic classification. Thus, with regard to the pessimistic procedure, it is verified that the BPMN was assigned to the category of Most Suitable. The UML and EEML languages were assigned to the category of Possibly Suitable. Thereafter, he/she conducted a sensitivity analysis and verified the robustness of the classification.

As just one modelling language was assigned to the category of Most Suitable, the manager should select this language since only that language meets his/her modelling objectives (Steps 5 and 6).

The fact of BPMN having been selected in this paper has no relation to BPMN having been chosen in the paper by Nysetvold and Krogstie (2005), because although both papers use the same table to show how the modelling languages were evaluated, each used a different procedure to aggregate data in order to arrive at a decision. In this present study, we suggest the use of a formal method that enables the decision be made in rational and efficient way.

7. Discussion

The main contribution of this paper is to propose a framework for helping the manager/ modeller to evaluate and select a modelling language by means of consistent methods.

The framework proposed in this study can evaluate any number of modelling languages, since neither the quality framework nor the ELECTRE TRI-B method put restrictions on the number of alternatives to be analysed.

Another advantage of using the ELECTRE TRI-B multicriteria method for classifying modelling languages and helping select one is that it enables the manager/modeller to make a decision that simultaneously takes into account several criteria which are often conflicting or may even be subjective. Furthermore, the ELECTRE TRI-B method helps him to organise and analyse the information and thus feel more confident in making a decision.

The SEQUAL is suggested for evaluating the business process modelling language because it considers the aspects most relevant to the quality of modelling languages, including physical, empirical, syntactic, semantic, pragmatic, social and organisational qualities (Krogstie, Sindre, and Jørgensen 2006). This feature enables the modelling language to be evaluated according to the wide range of criteria needed to investigate the appropriateness of the modelling language for conducting the modelling (Cortes- Cornax et al. 2011).

It would not be possible to achieve this type of evaluation with the BWW framework, since it focuses just on the conceptual basis of a modelling language (Wahl and Sindre 2005). In other words, the BWW framework concentrates on analysing the capabilities of the modelling language to give complete and clear descriptions of the domain being modelled by means of mapping constructs in the representational model and the model- ling language (Indulska et al. 2008). Nor would the workflow patterns framework do because it establishes several patterns and investigates the capacity of the modelling language in expressing each of these patterns. The modelling languages are evaluated only in terms of control flow, data, resource and exception handling perspective. Elements such as user-friendliness and ease of learning are not examined by the workflow patterns framework (Wohed et al. 2009).

As mentioned earlier, it is necessary to define criteria related to each of the six areas of language quality established by the quality framework in order to evaluate the modelling

28 A.C. Scanavachi Moreira Campos and A.T. de Almeida

languages. This paper is based on what Nysetvold and Krogstie (2005) proposed and use was made of a list of criteria defined by Østbø (2000). Nevertheless, any other means of establishing criteria can be used, provided that such criteria be complete (exhaustive), non-redundant, concise and can be measured either qualitatively or quantitatively.

Although it is not easy to make quantitative evaluation using the SEQUAL, its focus on qualitative evaluations enables the quality of modelling languages to be evaluated (Krogstie, Sindre, and Jørgensen 2006). Moreover, the use of the ELECTRE TRI-B multicriteria method enables the uncertainties in the modelling language evaluations to be taken into account by means of defining values for preference, indifference and veto thresholds. Consequently, this method will assign modelling languages to different cate- gories only if they are really different.

At the same time, the need to elicit several parameters (preference, indifference and veto thresholds, weights, reference profiles and cutting level) in order to build a preference model of the DM is a disadvantage of the ELECTRE TRI-B method. Accordingly, the presence of an analyst is required in order to help the DM to structure the problem and quantify the DM’s preferences. However, there are some techniques that can be used by the analyst in order to elicit the value of parameters from the DM, as for example, the one presented by Mousseau and Slowinski (1998) that proposes eliciting the parameters of the model from assignment examples.

8. Conclusion and further research

This paper proposes a framework for aiding the manager/designer to evaluate and select a modelling language for BPM according to his/her modelling purposes based on the SEQUAL and ELECTRE TRI-B multicriteria method. This study makes no attempt to establish new forms of assessment and evaluation criteria, but rather intends to demon- strate how two existing approaches can be combined to solve the problem of selecting a modelling language. The advantages and disadvantages of using these approaches in an integrated manner were discussed.

Therefore, the current paper offers a scientific contribution to the management field, particularly considering the enterprise information system, since this study proposes the integration of knowledge from two different areas, thus facilitating the improvement in the context of process selection for a business process modelling language that occur in various organisations in the context of real-world application of BPM concepts.

By using the proposed framework, the manager/modeller can identify among all modelling languages those that best meet his/her modelling purposes by means of classifying them. After doing so, the manager/modeller can deepen his/her knowledge in those languages classified as most appropriate for conducting the modelling and then make a more adequate choice. This framework is demonstrated by means of an example.

A limitation of this study is related to defining criteria associated with each of the six areas of language quality defined by the quality framework. We use the list of criteria defined by Østbø (2000) as Nysetvold and Krogstie (2005) proposed, but we recognise that further studies should be undertaken in this field so that an exhaustive list of evaluation criteria can be developed.

A separate study is needed on the possibility of having models for group decision processes, especially those connected to collaborative BPM.

Enterprise Information Systems 29

References Aagesen, G., and J. Krogstie. 2010. “Analysis and Design of Business Process Using BPMN.” In

Handbook on Business Process Management 1-Introduction, Methods and Information Systems, edited by J. Vom Brocke and M. Rosemann, 213–235. Berlin: Springer-Verlag.

Aguilar-Savén, R. S. 2004. “Business Process Modelling: Review and Framework.” International Journal of Production Economics 90: 129–149. doi:10.1016/S0925–5273(03)00102–6.

Belton, V., and T. J. Stewart. 2002. Multiple Criteria Decision Analysis: An Integrated Approach. Boston, MA: Kluwer Academic Publishers.

Brito, A. J., A. T. de Almeida, and C. M. M. Mota. 2010. “A Multicriteria Model for Risk Sorting of Natural Gas Pipelines Based on ELECTRE TRI Integrating Utility Theory.” European Journal of Operational Research 200: 812–821. doi:10.1016/j.ejor.2009.01.016.

Brito, A. J., A. T. de Almeida-Filho, and A. T. de Almeida. 2010. “Multi-Criteria Decision Model for Selecting Repair Contracts by Applying Utility Theory and Variable Interdependent Parameters.” IMA Journal of Management Mathematics 21: 349–361. doi:10.1093/imaman/ dpn014.

Campos, A. C. S. M., S. F. D. Daher, and A. T. de Almeida. 2011. “New Patents on Business Process Management Information Systems and Decision Support.” Recent Patents on Computer Science 4: 91–97. doi:10.2174/2213275911104020091.

Capozucca, A., and N. Guelfi. 2010. “Modelling Dependable Collaborative Time-Constrained Business Processes.” Enterprise Information Systems 4 (2): 153–214. doi:10.1080/ 17517571003753266.

Chu, X. N., S. K. Tso, W. J. Zhang, and Q. Li. 2002. “Partnership Synthesis for Virtual Enterprises.” The International Journal of Advanced Manufacturing Technology 19: 384–391. doi:10.1007/ s001700200028.

Cortes-Cornax, M., S. Dupuy-Chessa, D. Rieu, and M. Dumas. 2011. “Evaluating Choreographies in BPMN 2.0 Using an Extended Quality Framework.” Lecture Notes in Business Information Processing 95: 103–117. In: Third International Workshop, BPMN 2011, Lucerne, November 21–22.

de Almeida, A. T. 2007. “Multicriteria Decision Model for Outsourcing Contracts Selection Based on Utility Function and ELECTRE Method.” Computers and Operations Research 34: 3569– 3574. doi:10.1016/j.cor.2006.01.003.

Elzinga, D. J., T. Horak, C. Y. Lee, and C. Bruner. 1995. “Business Process Management: Survey and Methodology.” IEEE Transactions on Engineering Management 42 (2): 119–128. doi:10.1109/17.387274.

Figl, K., J. Mendling, and M. Strembeck. 2009. “Towards A Usability Assessment of Process Modeling Languages.” CEUR Workshop Proceedings 554: 118–136. In: Proceedings of the 8th Workshop Geschäftsprozessmanagement Mit Ereignisgesteuerten Prozessketten (EPK 2009), Berlin, November 26–27.

Figueira, J., S. Greco, and M. Ehrgott, eds. 2005. Multiple Criteria Decision Analysis: State of the Art Surveys. New York: Springer.

Giaglis, G. M. 2001. “Taxonomy of Business Process Modeling and Information Systems Modeling Techniques.” The International Journal of Flexible Manufacturing Systems 13: 209–228. doi:10.1023/A:1011139719773.

Glassey, O. 2008. “A Case Study on Process Modelling – Three Questions and Three Techniques.” Decision Support Systems 44 (4): 842–853. doi:10.1016/j.dss.2007.10.004.

Indulska, M., J. Recker, P. Green, and M. Rosemann. 2008. “Representational Deficiency of Process Modeling Languages: Measures and Implications.” In Proceedings of the 16th European Conference on Information Systems, edited by W. Golden, T. Acton, K. Conboy, H. van der Heijden, and V. Tuunainen, June 9–11, 1632–1643. Ireland: ECIS.

Krogstie, J. 2003. “Evaluating UML Using A Generic Quality Framework.” In UML and the Unified Process, edited by L. Favre, 1–22. Hershey, PA: Idea Group Publishing.

Krogstie, J., and S. Arnesen. 2005. “Assessing Enterprise Modeling Languages Using a Generic Quality Framework.” In Information Modeling Methods and Methodologies: Advanced Topics in Database Research, edited by J. Krogstie, T. Halpin, and K. Siau, 63–79. IGI Global. doi:10.4018/978-1-59140-375-3.ch004.

Krogstie, J., G. Sindre, and H. Jørgensen. 2006. “Process Models Representing Knowledge for Action: A Revised Quality Framework.” European Journal of Information Systems 15: 91–102. doi:10.1057/palgrave.ejis.3000598.

30 A.C. Scanavachi Moreira Campos and A.T. de Almeida

La Rosa, M., A. H. M. ter Hofstede, P. Wohed, H. A. Reijers, J. Mendling, and W. M. P. van der Aalst. 2011. “Managing Process Model Complexity Via Concrete Syntax Modifications.” IEEE Transactions on Industrial Informatics 7 (2): 255–265. doi:10.1109/TII.2011.2124467.

Li, Y., B. Cao, L. Xu, J. Yin, S. Deng, Y. Yin, and Z. Wu. 2013. “An Efficient Recommendation Method for Improving Business Process Modeling.” IEEE Transactions on Industrial Informatics 10 (1): 502–513. doi:10.1109/TII.2013.2258677.

Liu, X., W. J. Zhang, R. Radhakrishnan, and Y. L. Tu. 2008. “Manufacturing Perspective of Enterprise Application Integration: The State of the Art Review.” International Journal of Production Research 46 (16): 4567–4596. doi:10.1080/00207540701263325.

Lourenço, R. P., and J. P. Costa. 2004. “Using ELECTRE TRI Outranking Method to Sort MOMILP Nondominated Solutions.” European Journal of Operational Research 153: 271–289. doi:10.1016/S0377-2217(03)00150-4.

Luo, W., and Y. Tung. 1999. “A Framework for Selecting Business Process Modeling Methods.” Industrial Management & Data Systems 99: 312–319. doi:10.1108/02635579910262535.

Merad, M. M., T. Verdel, B. Roy, and S. Kouniali. 2004. “Use of Multi-Criteria Decision-Aids for Risk Zoning and Management of Large Area Subjected to Mining-Induced Hazards.” Tunnelling and Underground Space Technology 19: 125–138. doi:10.1016/S0886-7798(03) 00106-8.

Morais, D. C., and A. T. de Almeida. 2006. “Water Supply System Decision-Making Using Multicriteria Analysis.” Water SA 32 (2): 229–235.

Mota, C. M. M., and A. T. de Almeida. 2012. “A Multicriteria Decision Model for Assigning Priority Classes to Activities in Project Management.” Annals of Operations Research 199: 361–372. doi:10.1007/s10479-011-0853-z.

Mota, C. M. M., A. T. de Almeida, and L. H. Alencar. 2009. “A Multiple Criteria Decision Model for Assigning Priorities to Activities in Project Management.” International Journal of Project Management 27 (2): 175–181. doi:10.1016/j.ijproman.2008.08.005.

Mousseau, V., and R. Slowinski. 1998. “Inferring An ELECTRE TRI Model from Assignment Examples.” Journal of Global Optimization 12: 157–174. doi:10.1023/A:1008210427517.

Nysetvold, A. G., and J. Krogstie, 2005. “Assessing Business Processing Modeling Languages Using A Generic Quality Framework.” In Proceedings of the 10th Workshop on Evaluating Modeling Methods for Systems Analysis and Design (EMMSAD’05), held in conjunction with the 17th Conference on Advanced Information Systems (CAiSE’05), edited by T.A. Halpin, K. Siau, and J. Krogstie, Porto, June 13–17, 545–556, FEUP.

Østbø, M. 2000. “Anvendelse av UML til dokumentering av generiske systemer (In Norwegian).” Master Thesis, Høgskolen i Stavanger, Norway.

Perçin, S. 2010. “Use of Analytic Network Process in Selecting Knowledge Management Strategies.” Management Research Review 33 (5): 452–471. doi:10.1108/01409171011041893.

Quartel, D., M. W. A. Steen, and M. M. Lankhorst. 2012. “Application and Project Portfolio Valuation Using Enterprise Architecture and Business Requirements Modelling.” Enterprise Information Systems 6 (2): 189–213. doi:10.1080/17517575.2011.625571.

Recker, J., M. Rosemann, M. Indulska, and P. Green. 2009. “Business Process Modeling: A Comparative Analysis.” Journal of the Association for Information Systems 10 (4): 333–363.

Roy, B. 1996. Multicriteria Methodology for Decision Aiding. Netherlands: Kluwer Academic. Silva, V. B. S., D. C. Morais, and A. T. de Almeida. 2010. “A Multicriteria Group Decision Model

to Support Watershed Committees in Brazil.” Water Resources Management 24: 4075–4091. doi:10.1007/s11269-010-9648-2.

Söderström, E., B. Andersson, P. Johannesson, E. Perjons, and B. Wangler, 2002. “Towards A Framework for Comparing Process Modeling Languages.” In 14th International Conference on Advanced Information Systems Engineering, CAiSE 2002, Canada, May 27–31. Lecture Notes in Computer Science, 2348, 600–611.

Su, X., and L. Ilebrekke, 2002. “A Comparative Study of Ontology Languages and Tools.” In 14th International Conference on Advanced Information Systems Engineering, CAiSE 2002, Canada, May 27–31. Lecture Notes in Computer Science, 2348, 761–776.

Tan, W., W. Shen, L. Xu, B. Zhou, and L. Li. 2008. “A Business Process Intelligence System for Enterprise Process Performance Management.” IEEE Transactions on SMC Part C 38 (6): 745–756.

Tan, W., W. Xu, F. Yang, L. Xu, and C. Jiang. 2013. “A Framework for Service Enterprise Workflow Simulation with Multi-Agents Cooperation.” Enterprise Information Systems 7 (4): 523–542. doi:10.1080/17517575.2012.660503.

Enterprise Information Systems 31

van der Aalst, W. M. P., A. H. M. ter Hofstede, B. Kiepuszewski, and A. P. Barros. 2003. “Workflow Patterns.” Distributed and Parallel Databases 14 (1): 5–51. doi:10.1023/ A:1022883727209.

Vergidis, K., C. J. Turner, and A. Tiwari. 2008. “Business Process Perspectives: Theoretical Developments Vs. Real-World Practice.” International Journal of Production Economics 114: 91–104. doi:10.1016/j.ijpe.2007.12.009.

Vincke, P. 1992. Multicriteria Decision-Aid. London: John Wiley. Viriyasitavat, W., L. D. Xu, and A. Martin. 2012. “Swspec: The Requirements Specification

Language in Service Workflow Environments.” IEEE Transactions on Industrial Informatics 8 (3): 631–638. doi:10.1109/TII.2011.2182519.

Wahl, T., and G. Sindre, 2005. “An Analytical Evaluation of BPMN Using a Semiotic Quality Framework.” In Proceedings of the Workshop on Evaluating Modeling Methods for Systems Analysis and Design (EMMSAD’05), held in conjunction with the 17th Conference on Advanced Information Systems (CAiSE’05), edited by T. A. Halpin, K. Siau, and J. Krogstie, Porto, June 13–17, 533–544. FEUP.

Wand, Y., and R. Weber. 1993. “On the Ontological Expressiveness of Information Systems Analysis and Design Grammars.” Information Systems Journal 3: 217–237. doi:10.1111/ j.1365-2575.1993.tb00127.x.

Weske, M. 2007. Business Process Management: Concepts, Languages, Architectures. Berlin: Springer-Verlag.

Wohed, P., N. Russell, A. H. M. ter Hofstede, B. Andersson, and W. M. P. van der Aalst. 2009. “Patterns-Based Evaluation of Open Source BPM Systems: The Cases of Jbpm, Openwfe, and Enhydra Shark.” Information and Software Technology 51 (8): 1187–1216. doi:10.1016/j. infsof.2009.02.002.

Xu, L. 2011. “Enterprise Systems: State-Of-The-Art and Future Trends.” IEEE Transactions on Industrial Informatics 7 (4): 630–640. doi:10.1109/TII.2011.2167156.

Xu, L., H. Liu, S. Wang, and K. Wang. 2009. “Modelling and Analysis Techniques for Cross- Organizational Workflow Systems.” Systems Research and Behavioral Science 26 (3): 367–389. doi:10.1002/sres.978.

Xu, L., W. Tan, H. Zhen, and W. Shen. 2008. “An Approach to Enterprise Process Dynamic Modeling Supporting Enterprise Process Evolution.” Information Systems Frontiers 10 (5): 611–624. doi:10.1007/s10796-008-9114-3.

Xu, L., W. Viriyasitavat, P. Ruchikachorn, and A. Martin. 2012. “Using Propositional Logic for Requirements Verification of Service Workflow.” IEEE Transactions on Industrial Informatics 8 (3): 639–646. doi:10.1109/TII.2012.2187908.

Zopounidis, C., and M. Doumpos. 2002. “Multicriteria Classification and Sorting Methods: A Literature Review.” European Journal of Operational Research 138: 229–246. doi:10.1016/ S0377-2217(01)00243-0.

32 A.C. Scanavachi Moreira Campos and A.T. de Almeida

Copyright of Enterprise Information Systems is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.