Construction Model Generation Based on DEMO: Case Study of Telecommunication Industry in Indonesia

Enterprise engineering is a discipline aspect of an enterprise, including designing and modeling a system. To create a new system, we can perform manipulation of existing systems. A construction system can be decomposed into several subsystems, and those subsystems can be merged into another construction system. Construction of an enterprise can be represented by DEMO (Design & Engineering Methodology for Organizations), specifically DEMO Construction Model. This research attempts to demonstrate the process of construction model generation using DEMO, and also develops the criteria for submodels and merged models. This article applies a case study of telecommunication industry in Indonesia. By using this framework, one can gather several enterprise models and model them as construction models, generate submodels and create a pool of submodels, merge them to generate a new construction model.


Introduction
The competitive nature of business in this modern society leads to the necessity of business innovation to create a new business. To create a new, successful business, a deep and complete understanding of ontological aspect of enterprise as a system is necessary, aided by Enterprise Engineering. Enterprise Engineering (EE) is a discipline aspect of an enterprise, including designing and modeling a system [1]. Three major disciplines of EE are Enterprise Architecture, Enter-How to cite this paper: Pratama, N.R. and struction of enterprises.
In order to conduct (re)design of an enterprise, an enterprise model is used to perform the (re)design prior to the implementation. To create a new system, we can perform manipulation of the construction of existing systems. Construction system of an enterprise can be represented by DEMO (Design & Engineering Methodology for Organizations) [2], and, to date, the only methodology available in EO. In particular, DEMO Construction Model as one of the aspect models of DEMO can be decomposed into several subsystems, and those subsystems can be merged into another construction system [3]. The manipulation, merging and decomposing DEMO Construction Model can be explained in algebraic notation [4], therefore it is possible to create a pool of submodels of construction models, and then merging them to create a new construction model. Figure 1 illustrated the proposition.
To create a new construction model, we can modify the existing one by a process of (de)composition of a model, called manipulation, in line with enterprise engineering concept. A manipulation is performed using a formal set of rules. DEMO Construction Model can be manipulated, merged, or decomposed by using algebraic notation based on prior research ( [3], further explained in [4]), making it possible to analyze and synthesize, or simply manipulate, construction models. They defined a model, a submodel as a part of the model, and the three basic operations (merge, complement, and digest). Given a global model, one can construct the family of all the submodels and the three basic operations.
They also showed that those operations are closed on the domain, and that some are commutative and associative. The advantage of this method is removing the necessity of model checking. A tool support for DEMO Construction Model was also developed [5], to provide functions to execute DEMO Construction Model manipulation without manual calculation.
Although the framework of DEMO Construction Model manipulation and tool support has been established, the practical application of this approach has not been intensively discussed. The ultimate objective of this research field is to minimize the aspect of craftsmanship in enterprise design, and bring enterprise engineering closer to design and/or engineering work for realizing agile enterprises [4]. However, the prior research lacks explanation of the criteria of submodel, related to the proposition of creating a pool of submodels and then merging them to create a new model. This leads to the following problems: • The practicality of this approach is not verified.
• There is no clear definition or criteria of the submodel split from the existing model. • The criteria for the applicable merged models also not yet established.
Hence, to address those problems, this study aims to demonstrate the process of construction model generation using DEMO Construction Model, and also develops the criteria for submodels and merged models. This article applies a case study of telecommunication industry in Indonesia, because there can be many construction models that can be captured in this industry. In Indonesia, business of telecommunication industry has developed rapidly [6]. Telecommunication companies in Indonesia have provided the best services, improving operational systems to improve their business performance [7]. [8]  This work will serve as a method of construction model generation and manipulation. By using this framework, one can gather several enterprise models and model them as construction models, generate submodels and create a pool of submodels, merge them to generate a new construction model in a certain industry.
This article aims to answer these following research questions to address the emerged problems: 1) How can we perform Construction Model Generation of Telecommunication industry?
2) What are the criteria of submodels split from the existing model?
3) What are the criteria for merged models to be applicable?
The remainder of this article is composed as follows. Section 2 will explain the literature review related to this work. Section 3 will explain the phase of the framework and proposed phases taken in each phase. The application of this framework in telecommunication industry will be mentioned in Section 4. Section 5 will cover the discussion of the result and the conclusion of this research will be given in Section 6. Journal of Service Science and Management Transaction kinds in DEMO CM may be related to each other. It is possible that a transaction kind, in practice, is inside the flow of another transaction kind, or in other words, belongs to a Transaction Tree [2]. Taking

Manipulation of DEMO Construction Model
DEMO Construction Model, with ATD as its representation, emphasizes clear coordination models so that it is possible to conduct manipulation of CM. In the literature, there exist several researches of DEMO application to solve practical problems, such as project [16], production [17], and even emergency management [18]. Among those researches, a series of research [19] [20] provides general guidelines of enterprise splitting and allying, using real cases with DEMO model, and tests the validity of designing enterprises in the real world via DEMO models.
[4] developed a formal specification of ATD and its submodels and defined algebraic operation as a means of model manipulation. They formalized the definition of a transaction kind T, an actor role A, a model

A T A T that is in both submodels.
They also showed that those operations are closed on the domain, and that some are commutative and associative. The advantage of this method is removing the necessity of model checking. A tool support for DEMO Construction Model was also developed [5], to provide functions to execute DEMO Construction Model manipulation without manual calculation. This tool is designed to automate the conversion between ATD of DEMO CM and mathematical objects, and the execution of computation so that users can input the CMs, perform manipulation operators, and then retrieve the results. Using this tool, users can focus on the manipulation itself rather than mathematical computation and translation.

Methodology
This section explains the framework of Construction Model Generation using split and merge operation to demonstrate the case study. The framework is illustrated in Figure 4.
To better understand this section, we introduced some related statements below. To understand deeply about construction of algebra, see [4]:   2) There is a set E consists of environmental actor roles, and a set B consists of border transaction kinds. For example in Figure 5, the environmental actor roles in α is a submodel α 1 corresponds to model α consists of actor roles A 1 and A 2 , and transaction kinds T 1 and T 2 , as depicted in Figure   5.

5)
In the split operation, we remove the system boundary line that defines the Scope of Interest (SoI) for the sake of simplicity in the specification of algebraic notation. In merge operation, we draw the system boundary based on the attribute of actor roles and transaction kinds that composed the merged model.

PHASE 1: Split Operation of the Initial CM to Create Submodels
In this phase, one or several initial models are split into several submodels. To maximize the effectiveness of the overall process, a submodel must also comply with some defined criteria. In addition to conditions stated in 3), a submodel resulting from split operation must also hold the following criteria: 1) Contains at least 1 transaction kind T and corresponding actor roles (initiator and executor) (A, A'). 4) The submodels from the same given model should be mutually exclusive with each other, i.e. the digest of those submodels results in an empty set a) Payment-Coupled Submodel: Parent-Child Transaction submodel that involves "payment" transaction kinds from a "customer" actor role denoted as p-type submodel; b) Border Submodel: A submodel that has at least one border transaction kinds that do not involve "payment" transaction from a "customer" actor role denoted as q-type submodel; c) No-border Submodel: A submodel that has no border transaction kinds denoted as r-type submodel.
If a submodel is able to be classified as Payment-Coupled Submodel, then it cannot be classified as Border Submodel or No-border Submodel, making the classification mutually exclusive with each other. Looking at Figure 4, if we assume that T 2 is Payment Transaction, then submodel α 1 is Payment-Coupled Submodel, submodel α 2 is Border Submodel, and submodel α 3 is a No-border Submodel.
To split a submodel from a model, we can follow these steps: Step 1: Select pair of actor roles and transaction kinds complies with the Criteria 1, 2, and 3, and determine the complement of the submodel.
Step 2: Check the resulting complement whether it complies with the Criteria 1, 2, and 3. If it is, the complement also becomes a submodel. If not, repeat Step 1 using the resulted component.
The repeating process in Step 2 complies with the Criteria 4. Figure 5 also shows the split operation of initial model α. As mentioned, submodel in α is a submodel α 1 corresponds to initial model α.
After the submodels of several initial models are generated, a pool of submodels can be created.

PHASE 2: Merge Operation of Submodels
To find the possibility of a new model that is meaningful, the following criteria of a new model must be fulfilled: 1) There can be only one p-type submodel in a new model. A new model is meant to be a simple business model, therefore only one payment transaction is necessary.
2) New model cannot have r-type submodels. The new model focused on transaction between internal and environmental actor roles, therefore we exclude submodel without border transaction kinds.
3) The new model is composed of no more than 3 submodels. More combination of submodels means expanding the possibility of new models; too many combinations can make the new model selection become too complicated. 4) New model cannot be an element of a single initial model ( * ⊆ A A and Journal of Service Science and Management * ⊆ T T ), as it will not be a new model if it is only a part of one initial model.
Using these criteria, new model consists of one p-type submodel and up to two q-type submodels. The possibility of new models depends on the number of combination of p-type and q-type submodels. If necessary, preliminary selection of submodels to be included in merge operation can be conducted to narrow down the possibility of non-meaningful models. Of course, it is possible to manually select those submodels and merge them to create a new, meaningful model. After the models are merged, draw system boundary line based on the attributes of actor roles and transaction kinds.

PHASE 3: Selection of New Models
After

Case Study: Telecommunication Industry in Indonesia
This section demonstrates the case study of Telecommunication Industry in Indonesia. Telecommunication industry, or sector, consists of companies that makes communication possible through phone or Internet, allowing data in words, voice, audio or video to be sent anywhere in the world [21]. In this sec-

DEMO Construction Model of Indonesian Telecommunication Industry
In this study, we captured six DEMO Construction Models as initial models from three Indonesian telecommunication companies. The CMs are captured from the existing business elements of the company, and represents telecommunication sector defined in [21]. The CMs captured are: For simplicity, this paper will only explain one of the captured CM: Mobile Internet Package, as one of the main business of Telecommunication Industry.

Mobile Internet Package
This is the standard mobile Internet package, to provide guaranteed mobile internet service. The TPT and ATD of Mobile Internet Package can be seen in Table 1 and Figure 6 respectively. The following is the description of Mobile Internet Package: A mobile user who wants to use Internet service can buy or subscribe to the Internet package from designated outlets, or via USSD service with the help of the contact center. The company develops Internet products, conduct promotion, and activate the Internet service via a network system. Billing system deducts the amount of Internet service fee from the mobile balance (prepaid), or via credit card (postpaid).
The ATD represents a CM of Mobile Internet Package. Note that the numbering of Actor Roles complies with Condition 3 in Section 2.2; the executor of a transaction kind gets the same number as the transaction kind, regardless of actor role types.

Construction Model Generation of Telecommunication Industry
In this section, we applied the phases introduced in Section 3.

PHASE 1: Split Operation of the Initial CM to Create Submodels
The first phase of this phase is to split the initial CM to create submodels. Given the initial CM of Mobile Internet Package α M , we can create submodels using the steps introduced in Section 3.1. The submodels of Mobile Internet Package is illustrated in Figure 7, complies with the criteria introduced in Section 3.1.
The submodels denoted as follows: After we conduct split operation in the entire initial model, we can summarize it in Table 2.

PHASE 2: Merge Operation of Submodels
Based on Criteria 1, 2, and 3 of Section 3.2, we can calculate the number of   possibility of the new merged models, by calculating the combination of p-type and q-type submodels satisfying the conditions of a new model, minus the merged models that have the exact same composition with the submodel of initial models. The total number of possible merged model with 1 P and up to 2 Q, where P is the number of p-type submodels and Q is the number of q-type submodels, is equal to: with 8 p-type submodels and 8 q-type submodels, the number is ( ) In total 24 models, subtracting 296 by 24 we found the possibility of new models is 272 models.

PHASE 3: Selection of New Models
In previous phases, we found the possibility of new models is 272 models. This number is too large to check one by one, so we examine the submodels to narrow down the possibility of new models. We determined that 4 of the submodels (M α2 , M α3 , M θ2 , and M θ4 ) are a generic submodel that does not change the essence of the business, and M θ4 is identical to M δ2 , therefore M α2 , M α3 , M θ2 , M θ4 , and M θ4 are excluded. Table 3 summarizes the submodels that are used in the next merge operation.
To further narrow down the possibility of new models that are meaningful, we can conduct cluster analysis to determine the group of the submodels so that submodels that belong to different groups cannot be merged (a submodel belongs to more than one group is possible). Submodels of the same initial model will belong to the same group, at least in one of the groups. By examining the relative closeness 1 between each submodel, we determined that there are two groups, with the summary illustrated in Table 4. Group 1 is a cluster of submodels related to financial transactions, and Group 2 is a cluster of submodels related to entertainment.
Using Table 4, we can calculate the number of combination of merged submodels for each group. In total 11 models. Subtracting 61 by 11, we found the possibility of the new models is 50 models instead of 272 models. This number is still relatively large; 1 If the two submodels merged will potentially results a meaningful model, those models are relatively close.  Table 4. Groups of Submodels.       Therefore it can be said that the framework is successful to synthesize a new enterprise model from the existing models. However, there are limitations regarding the newness of the models. Using the criteria of applicable merged models we avoid any possible recurrence of the initial models, however, there is still no way to avoid recurrence of business models outside of the captured models.

Discussion
Therefore to minimize this "fake" newness, existing enterprise models in an industry have to be captured as many as possible. This way we can also expand the pool of submodels, therefore the possibility of new, meaningful models can be improved. We can see that using DEMO in model manipulation has its merit and demerit. The advantage of DEMO is to provide the construction model that can be easily manipulated using algebraic notation. Manipulation of DEMO CM is proved to be closed on the domain, and that some are commutative and associative. The set of rules of models and submodels, and the manipulation operations, are all defined, so that all models and submodels are consistent with the definition of DEMO CM, removing the necessity of model checking. The disadvantage is the meaningfulness of the merged model needs to be discussed because there is no value-related context in DEMO. To address this, another modelling artefact containing value aspect can be introduced; one of the most popular is Business Model Canvas [22], commonly abbreviated as BMC, given that we can transform BMC into DEMO CM [23] or vice-versa [24].
By using the proposed methodology in this paper and future studies, we expect that we can gather several enterprise models as construction models and split them into several submodels and create a pool of submodels, merge them to generate a new construction model as a new enterprise model in a certain industry.

Conclusions
In this research, we identified some research problems related to construction model manipulation in Section 1 and explained the literature review related to this work in Section 2. We explained the framework of Construction Model This article demonstrated the Construction Model Generation as a method to synthesize a new enterprise model using Telecommunication Industry in Indonesia as a case study. We perform Construction Model Generation using case study of Telecommunication Industry to demonstrate the model manipulation using DEMO Construction Model. We defined the criteria of submodels split from the existing model in. We also defined the criteria of merged models to be applicable, so that these models can potentially be implemented as a new enterprise model. Therefore, we answered all three research questions. The first question is described in Section 3, enhanced with the case study of Telecommunication Industry in Indonesia described in Section 4, completed with the results. The second question is described in Section 3.1, we developed 4 criteria and 3 classifications of submodel to help the manipulation process. The third question is described in Section 3.2, we developed 4 criteria of merged model to narrow down the possibility of new models, and eliminate merged models that are not applicable or recurrence of an initial model.
In Telecommunication Industry in Indonesia, we captured 6 existing enterprise models as DEMO Construction Model. We conduct split and merge operation of those models, and we found 3 new, meaningful models of DEMO Construction Model as a new enterprise model that is meaningful and applicable in Telecommunication Industry. The advantage of DEMO is to provide the construction model that can be easily manipulated using algebraic notation. The set of rules of models and submodels, and the manipulation operations, are all defined, so that all models and submodels are consistent with the definition of DEMO, removing the necessity of model checking. The disadvantage is the meaningfulness of the merged model needs to be discussed, because there is no value-related context in DEMO; the authors propose using Business Model Canvas to extend the framework and add the value aspect to the framework. By using this methodology, we expect that we can gather several enterprise models as construction models and split them into several submodels and create a pool of submodels, merge them to generate a new construction model.