Business Process Model Recommendation as a Transformation

Process in MDE: Conceptualization and First Experiments

Hadjer Khider

1,2 a

, Slimane Hammoudi

and Abdelkrim Meziane

Information Systems and Multimedia Systems Department, CERIST, Algiers, Algeria

Computer Science Department, Faculty of Exact Sciences, University of Bejaia, Bejaia 06000, Algeria

ERIS Team, Computer Science Department – ESEO, Angers, France

Keywords: Business Process Model Reuse, User Business Profile, Business Process Modeling, Recommender System,

MDA, Meta Models, Transformation, Weaving, Recommendation Business Process Model.

Abstract: Business Process (BP) model repositories have been proposed to store models of BP and make them available

to their stakeholders for future reuse. One of the challenges facing users of such repositories concerns the

retrieval of models that suit their business needs in a given situation, which is not provided by current

repositories. In order to overcome this lack, one important issue to investigate is to provide recommendation

of BP models based on the user profile as the most important way to better meet his business needs which

promote BP model reusability. In this paper, we propose a conceptual framework of BP model

recommendation based on the user social profile and implemented as a transformation process in model driven

engineering (MDE). In our experiments, the LinkedIn social network is used to extract the users’ business

interests. These user business interests are then used to recommend the appropriate BP models that could fit

to the user. Our proposed framework is based on model driven architecture (OMG MDE approach) where

techniques of models, metamodels, transformation and weaving are used to implement a generic

recommendation process.

1 INTRODUCTION

A Business Process (BP) modeling is fundamental

part of Business Process Management (BPM)

lifecycle for improving organizational efficiency and

quality of business processes (BPs) in enterprises

(Gerth, 2013). However, modeling BP from scratch is

fallible, complex, time-consuming and error prone

task (Markovic, Pereira, and Stojanovic, 2008). One

of the promising solutions to these issues is the reuse

of BP models (Elias and Johannesson, 2012).

Therefore, it is important to provide a repository to

store thousands of BP models for business reuse (Gao

and Krogstie, 2010).

Re-use techniques facilitate the use of existing BP

models in order to simplify the development of new

models or the improvement of existing BP models.

Reuse of BP models is designing BPs by using

existing process models (Gao and Krogstie, 2010).

Exploiting already designed BP models is one of the

promising solutions to reduce the time and cost

https://orcid.org/0000-0002-0566-9235

consumed to model new BP, minimize errors and

increase BP model quality and flexibility (Gao and

Krogstie, 2010), (Fellmann, Koschmider, and

Schoknecht, 2014), (Shahzad et al., 2009), (Mendling

et al., 2017). To reuse BP models, it is important to

provide a BP models repository to store BP models

for future reuse. Fortunately, many organizations

have establish BP model repositories to maintain and

reuse existing BP models (Elias and Johannesson,

2012),(Gao and Krogstie, 2010), (Fellmann,

Koschmider, and Schoknecht, 2014), (Shahzad et al.,

2009). For example, SAP, MIT.

Recent works (Elias and Johannesson, 2012),

(Shahzad et al., 2009), (Yan, Dijkman, and Grefen,

2017),(Yan and Grefen, 2010) show that the existing

repositories do not adequately support reuse of BP

models. So, it is difficult for business actors to find

the relevant BP models in BP model repositories and

share process knowledge. The lack of an efficient

models retrieval system (Yan, Dijkman, and Grefen,

2017), the different process repositories

classifications, the heterogeneity of repositories

Khider, H., Hammoudi, S. and Meziane, A.

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments.

DOI: 10.5220/0009155600650075

In Proceedings of the 8th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2020), pages 65-75

ISBN: 978-989-758-400-8; ISSN: 2184-4348

structure and the potential size of existing repositories

do not allow their users to find easily models for

future reuse. On the other hand, process model

retrieval from a process repository still suffer from

much manual work (Li et al, 2014) and cannot relieve

business users from the highly complex, time

consuming, and error-prone task of building new BPs

from scratch.

Considering the above issues, there is a need for

an automated approach which can recommend BP

models to help repository stakeholders to find and

therefore reuse models to build new one.

Recommender systems have been widely used as an

effective answer to these difficulties. Recently,

Recommender systems became a hot topic and have

been widely used in both academic research and

industry applications. However, its applicability in

the field of BPM remains very modest, there are only

a few works (Li et al., 2014), (Deng et al., 2017),

(Hornung, Koschmider, and Oberweis, 2009),(Imad,

Elkindy, and Corea, 2019), (Schonenberg et al.,

2008),(Setiawan, Sadiq, and Kirkman 2011)

(Laurenzi et al., 2019) in BPM domain.

We claim that by integrating recommender

system into BP model repository, we can increase the

accessibility of BP models for repository users.

Furthermore, integrating recommender system into

current BP model repositories presents an effective

mechanism that can help users, to find BP models

easily and reuse the available knowledge of BP model

repositories to reduce time and effort, and improve

the quality of newly designed models. Motivated by

this idea, SBPR, a Social Business Process

Recommender for BP model reuse, is proposed.

According to the study presented by Shahzad, et

al. in (Shahzad, Elias, and Johannesson, 2010), in

order to reuse BP models, it is important to know the

environment in which a process can or is intended to

work. This environment consists of the business

context in which the BP can be applied, the goals of

the process, and the actors of the process. We believe

that the business environment in which stakeholders

are involved can influence the BP models that may

interest them. The LinkedIn social network is used to

extract the business profiles of users. Research has

shown that LinkedIn is the primary business-oriented

social networking site that professionals use (Salman,

2019). LinkedIn had more than 630 million members

making it the largest online professional network

(Salman, 2019). A LinkedIn profile contains a

member’s current and past work history, education,

career information and projects which reflects the

business environment in which the users are actually

involved.

In this paper, we propose a novel recommender

system, which generates predictions simply by

combining the user business profile to determinate the

business environment in which user is actually

involved with the metadata of BP models stored in

repository. The power of our recommender system

lies in its ability to make a recommendation of BP

models in real time from users’ business profiles. Our

approach is based on model driven development

(MDD) through model driven architecture (MDA)

where metamodels allow to define a generic

representation of user profiles in one side and BP

models in other side. Taking advantage as much as

possible of existing MDA technologies, we have

implemented our recommender system as a

transformation process which takes as input a

LinkedIn user profile and proposes in output a set of

potential BP models.

In this paper, we propose a conceptual framework

of BP model recommendation to help users find new

BP models from repository in an efficient and

accurate way. The proposed framework is based on

the user social profile and implemented as a

transformation process in model driven architecture

(MDA).

The remainder of this paper is organized as

follows: Section 2 presents related work. Section 3

gives an overview of our SBPR framework. Section 4

presents our Model Driven Approach where

metamodels for both user business profile and BP

metadata are presented. Section 5 presents a case

study, results and discussion. Finally, section 6

concludes and outlines our future research directions.

2 RELATED WORK

Recommendation has already been investigated in

BPM to assist stakeholders in choosing the suitable

BP for a given activity. Presently, there are two kinds

of work on process model recommendation (Deng et

al., 2017):

- Complete process model recommendation by

reusing existing complete process models and

- Process model fragments recommendation in

order to model new process models by reusing

process model fragments.

2.1 Process Model Fragments

Recommendation for Modeling

A lot of work has been devoted to Process model

fragments recommendation. For example:

MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development

H. Schonenberg et al. in (Schonenberg et al.,

2008) proposed a recommendation service, which,

when used in combination with flexible PAISs

, can

support end users during process execution by giving

recommendations on possible next steps.

Recommendations are generated based on similar

past process executions by considering the specific

optimization goals.

Hornung et al. in (Hornung, Koschmider, and

Oberweis, 2009) have proposed a recommender

system that suggests a list of correct and fitting

process fragments for an edited BP model, which can

be used to complete the process model being edited.

The benefit of such a system is to assist users during

process modeling by reusing process fragments from

a process repository.

M. Andri Setiawan in (Setiawan, Sadiq, and

Kirkman, 2011) presented an approach to assist

process designers and promote process improvement

by the development of methods for multi-criteria

based process ranking and personalized

recommendation, the approach intended to assist

users by learning from already existing successful

practices.

Ying Li in (Li et al., 2014) proposed a workflow

recommendation for improving BP modeling that

precede associate rules mining between business

activity nodes and process fragments within the

workflow repository, and provide decision support

for modeling processes. The associate rules mining

refer to pattern extraction. Extraction of those

patterns from the workflow repository is based on

graph-mining technique. which is the foundation of

workflow recommendation. Based on measuring

similarity and calculating distance between reference

model and process patterns, technique could

recommend appropriate nodes for modelers to

automate the construction of BPs.

Rangiha et al. in (Rangiha, Comuzzi, and

Karakostas, 2015) described a recommender task

system that uses social tagging to collect relevant

information from discussions between process actors

during process execution. Analysis of these tags

allows the system for recommending new tasks when

the same process must be executed again.

Deng et al. in (Deng et al., 2017) have proposed a

process recommendation system that can assist BP

analysts build new BPs from scratch. It can

recommend proper nodes (fragments) based on

patterns mined from existing process repositories.

Process Aware Information Systems

E. Laurenzi et al. in (Laurenzi et al., 2019)

proposed a process modeling approach that assists

domain experts in the creation and adaptation of

process models. To get an appropriate assistance, the

approach is driven by semantic patterns and learning.

Semantic patterns are domain-specific and consist of

process model fragments (or end-to-end process

models), which are learned from feedback from

process modeling experts.

2.2 Complete Process Model

Recommendation

Complete process model recommendation helps to

reuse process models by discovering or retrieving

such process models from repository that satisfy

users’ explicit requirements or conform to their

implicit intentions.

A. Koschmider et al. in (Koschmider, Hornung,

and Oberweis, 2011) proposed a system for

supporting users at modeling time by providing a

recommender component and search functionality for

process model parts stored in a repository. The

recommendation-based modeling support system is

based on users’ tagging behavior and intentions, this

system implements other functionality, such as a

search interface.

K. Kluza et al. in (Kluza et al., 2013) presented

several machine learning methods which can be used

for recommending features of BP models.

Furthermore, the study in Kluza et al. (Kluza et al.,

2013) suggests a classification schema to the

recommendations.

Bobek et al. in (Bobek et al., 2013) presented a

method that uses Bayesian Networks for

recommendation purposes in process modeling and

configurable models, such a method can help in

speeding up modeling process and producing models

that are less error prone compared to these designed

from scratch.

Ariouat et al., in (Ariouat, Andonoff, and

Hanachi, 2018) has addressed process

recommendation in crisis management field. The

Process recommendation uses data observed in the

field, i.e. risk and damage of the crisis, along with

business knowledge of actors involved in crisis

resolution in order to recommend different strategies

with observed facts depending to the context and then

build processes corresponding to chosen strategies.

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments

Figure 1: SBPR Framework Source: (Khider et al., 2018).

2.3 Synthesis and Positioning

However, the above-cited approaches use

recommendation suggestions during design time and

usually focused on how to guide the modeler during

the modeling process on possible next activities to

create. Furthermore, those approaches ensure a

certain degree of modeling support but yet with little

focus on the user interests

. The recommendations

provided is based on factors as labels of elements,

current progress of modeling process, and additional

information, such as process description. but they

never take the business interests of the user into

consideration (Kluza et al., 2013). In this paper we

propose a process recommender system to help

business users find new process models from

repository in an efficient and accurate way.

Furthermore, our approach is not limited to designer

or modeling experts but it is destined to any user who

accesses a repository of process models in order to

find a BP model to reuse. Moreover, the power of our

recommender system lies in its ability to make

recommendation based on business profile of user

and implemented as a transformation process in

model driven architecture (MDA). Our proposition is

not limited to the modeling stage but it can be spread

on all stages of BPM. To our knowledge, there are no

approaches that have proposed the integration of a

recommender system to current process model

repositories to help their users find models easily and

according to their business needs.

Moreover, till now, according to our knowledge

there is no an approach that uses the MDA model

transformation process for the recommendation of BP

models. Thus, recommending the appropriate BP

models that could fit to the user would be

implemented as a transformation of a user business

profile into BPs models.

3 OUR SBPR FRAMEWORK

In this section, we propose the Social Business

Process Recommender (SBPR) Framework in order

to overcome some of problems that affect the actual

BP model repositories. The figure 1 illustrates the

different elements characterizing this framework and

their relationships. On the left hand side of this figure,

we find the main properties of a user profile according

to LinkedIn. On the right side, the metadata of BP

model are presented. SBPR recommender aims to

recommend to the users of process models

repositories BP models for reuse. LinkedIn User

profile is the source of social data for SBPR

recommender; BP models are target items to be

recommended to user. SBPR recommender

recommend BP models based on the business

environment of user which is learned from his

LinkedIn business profile.

MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development

4 A MODEL DRIVEN

APPROACH

In conformity with MDD approach we have defined

two metamodels, the first one is the LinkedIn user

profile metamodel and the second one is the BP

metadata metamodel.

4.1 LinkedIn User Profile Metamodel

LinkedIn is an

online

professional

social network

which may represent real-world professional

relationships

. We have chosen LinkedIn

for the

reason that LinkedIn is a business-oriented social

networking site where users post their professional

and career information, which allow us to have an

image of the user’s business environment. Figure 2

presents a potential meta-model for the LinkedIn User

profile (fragment).

The meta-model defined in this section was

designed as a synthesis of concepts proposed by

several authors, and more specifically the work of

Jean-Marie Favre presented in (Jean-Marie Favre

2007). The main elements of this metamodel are: (as

shown in figure 2)

 The LinkedIn User Profile: LinkedIn is used for

professional networking; it allows members to create

profiles. LinkedIn user profile is formed by a set of

profile fields

predefined in LinkedIn.

 LinkedIn User Business Profile: from the fields

that constitute the full user profile we have chosen

those that reflect his business environment. The main

elements that make up the business profile are:

Industry, Position, Skills.

 Industry: the user on LinkedIn can specify his

industry according to the reference table of industry

codes available on LinkedIn. the 10 top industries in

LinkedIn are: Information Technology and Services,

Marketing and Advertising, Human Resources,

Computer Software, Financial Services, Staffing and

Recruiting, Internet, Management Consulting,

Telecommunications, Retail

 Business Experience: includes information about

the business accomplishments, skills, company,

industry, area of expertise and professional

competencies of the user.

4.2 BP Metadata Metamodel

Figure 3 presents a possible meta-model for the BP

metadata (fragment). The BP metadata metamodel

(Figure 3) captures these elements and their

relationships in detail.

Figure 2: LinkedIn User profile metamodel.

https://www.linkedin.com/

https://developer.linkedin.com/docs/fields/full-profile

https://developer.linkedin.com/docs/reference/industry-

codes

https://www.linkedin.com/pulse/linkedin-industry-rank

ings-see-which-tops-list-joshua-waldman

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments

Figure 3: BP metadata metamodel (Fragment).

This metamodel is based on set of proposed

developed metamodels in literature among which the

metamodel developed by zur Muehlen in (Zur

Muehlen, 2004), the metamodel developed by M.

Rosemann (Rosemann, Recker, and Flender, 2008)

and metamodel developed by M.Elias in (Elias,

Shahzad, and Johannesson, 2010). The main elements

of this metamodel are:

 BP Description

Many organizations complement their BP models

with textual descriptions specify additional details

that describes the activities of a process, the involved

entities and their interaction. Process repositories in

practice do not only consist of BP models, but often

also contain textual BP descriptions (Leopold et al.,

2019). Taking the information from textual

descriptions into account may provide a clearer view

of the BP models that may interest the user in order

to reuse them. By taking into account the semantic of

textual descriptions of BP models in our

recommendation process allows us to identify more

relevant models from repository.

 Business Context

To reuse a BP model, it is important for a user to

understand the business environment in which a BP

is aimed to work. The business context in which the

BP takes place is specified by a set of categories and

their associates values (Hofreiter and Huemer, 2006).

e.g. Industry, Manufacturing. SAP repository

classifies BP models according to their business

context. According to Born et al. (Born, Kirchner, and

Müller, 2009), the context defines the environment in

which a BP is used.

 Business Goal

BP model describes activities conducted in order to

achieve business goals. Business goals express what

the organization wants to achieve from the business

perspective (Markovic, Pereira, and Stojanovic,

2008). As defined by M. Weske in (Weske, 2019) a

business goal is the target that an organization aims

to achieve by performing correctly, the related BP.

Relating BP models with goals in the repository can

help users to understand and thereby reuse BP

models.

4.3 A Methodology for Transformation

Process

In our proposed approach the transformation process

of a User business profile metamodel (source) into a

BP Metadata metamodel (target) aims to find

correspondences between the elements of two

metamodels source and target, and then generate

metadata to allow us to predict models that fit to

business environment of the user for reuse.

In conformity with the research work discussed in

(Hammoudi et al., 2010), the transformation process

is structured in two phases: mapping specification and

transformation definition.

A mapping specification is a definition of the

correspondences between two metamodels. From a

conceptual point of view, the explicit distinction

between mapping specification and transformation

definition remains in agreement with the MDA

philosophy, i.e. the separation of concerns.

Moreover, a mapping specification could be

MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development

associated with different transformation definitions,

where each transformation definition is based on a

giving transformation definition metamodel. Figure 4

hereafter represents a mapping specification from

LinkedIn User business profile into BP metadata. We

have used ATL transformation language (Bézivin et

al., 2003) to define transformation rules.

4.3.1 Generation of Transformation Rules

This phase aims to automatically generate

transformation rules from mappings, and format them

into a transformation models. According to Figure 4,

we have the following mappings:

Figure 4: Mapping from User business profile into BP

model metadata.

The User business profile business

environment is mapped into BP metadata

business context through the rule Be2Bc.

The User business profile business

experience is mapped into BP metadata business

textual description through the rule Be2T.

The User business profile Professional

Goal is mapped into BP metadata Business

Goal through the rule Pg2Bg.

The table 1 below show the different rules

generated.

Table1: Transformation Rules from mappings.

Transformation

rules in human

language

Transformation rules in ATL

R1 Be2Bc:

Business

environment

in LinkedIn

User business

Profile

corresponds to

business

context in BP

metadata

rule Be2Bc {

from be:

ser

usiness profile! business

environment to bCxt_

Grtd

metadata! business context(

context<- Be.category)

- - Cxt_

Grtd

: name of local

ariable referencing the

nstance created at the

output of the rule }

R2 Be2T:

Business

experience

(industry,

company,

position) in

LinkedIn User

Profile

corresponds to

textual

description in

BP metadata

rule Be2T {

from bxp: User

business profile! business

experience to

Des

_Grtd

BP Metadata!

Textual Description (

description<-

bxp.industry_title+bxp.pos

tion

title +

bxp.company_name)

- - Des

_Grtd

: name of

ocal variable referencing

he instance created at the

output of the rule}

R3 Pg2Bg:

Professional

Goal from

LinkedIn User

business

Profile

corresponds to

Business Goal

in BP

metadata

rule Pg2Bg {

from Pg: User

usiness profile!

professional goal

to Gl_

Grtd

Metadata! business goal (

BGoal<-pg.Goal)

- - Gl

_Grtd

: name of

ocal variable referencing

he instance created at the

output of the rule}

5 A FIRST EXPERIMENTS

5.1 Dataset Description

To test our approach, we have established dataset of

BP models and related metadata. The dataset we use

for applying our approach for recommendation of BP

Target

Transformation

Source

-Context: String

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments

models consists of 600 BP models modeled in the

Business Process Model and Notation (BPMN).The

collection of BP models is collected manually from

different BPM platform like Signavio

,Apromore

and other sources. In addition to the BP models, the

dataset contains 600 textual BP model descriptions in

the text format. Table 2 summarizes the key

characteristics of our dataset collection. The BP

models cover various business domain including

manufacturing process, selling process, booking

process …etc.

Table 2: Characteristics of dataset collection.

No. of BP models 600

Format

BPMN

No. of BP model textual

descriptions

600

No. of words (average)

No. of words / model

(average)

Format

64 800

108

.txt

5.2 Case Study

For the purpose of evaluation, we conduct a case

study to evaluate the effectiveness of our

recommender system by conducting experiments

applied to our dataset of 600 BP models in BPMN

format when user invokes a query about a claim

process. We have the following user: Patty Adel with

the following LinkedIn profile (view the table 3

below). In our case study the user enters ‘claim’ as a

keyword using the classic search functionality

available in repository to specify that he needs models

on claim process.

Table 3: LinkedIn user business profile.

Positions Title: Assistant, Business Development

positions/is-current: true

Positions/company-name: The Economical Insurance

Group

Public-profile-url:/pub/patty-adelvard-eig/39/90b/631

Location: Calgary, Canada Area

First-name: Patty

Last-name: Adel

Industry: Insurance

Skills: null

Current project: no specified

https://www.signavio.com/

https://apromore.org/

User inquiry (search keyword): claim

Probably results: for the user inquiry we have the

following results as shown in Table 4 below:

P70LR, P122HR,

P70NC, P11 ...etc. denotes BP

Models.

Table 4: Case study results.

User keyword claim

Repository collection Ours (dataset)

Process models founds

(by the classic search

functionality)

P70LR, P120LR, P116LR

P122, P120, P119, P116,

P113, P70, P11

P122HR, P119HR,

P116HR, P113HR, P70HR

P70NC, P122NC,

P120NC,113NC, P86NC

Number of the BP

models found

(classic search)

BP models format BPMN

Recommended BP

models according our

approach Social

recommendation + user

inquiry

P120LR

P120, P113

P113HR

P120 NC

P113NC

Number Of BP models

found via our

approach+ user inquiry

Total BP models in

dataset

600

5.3 Evaluation

To evaluate the performance of the proposed SBPR

recommender system in terms of Top-N

recommendations, we have chosen the following

metrics (Nadee, 2016) (i.e. Precision at N, Recall at N,

F1 measure) for the evaluation of the recommender.

We have chosen to evaluate the accuracy performance

of our recommender system in terms of top-5

recommendations. We ask recommender to provide

five BP models (we have chosen top 5 recommended

BP models).

We hide some BP models relevant to user U (our

user ‘Patty Adel’ case study) as a testing set.

Testing test= {

P120, P113, P113HR, P120 NC}

MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development

- Precision on top-5:

  

∩ /

- Recall on top-5:

  

∩ /

- F1 measure:

1  2    /



Table 5: Accuracy performance metrics computing for

SBPR recommender.

Accuracy performance

metrics (for SBPR

Recommender)

values

Precision

Recall

F1 measure

4/6=66.667 %

4/4=1=100 %

0.8

Figure 5: BP models found (results of the case study).

Figure 6: The accuracy performance comparison between

SBPR recommender and classic search functionality.

Table 6: Accuracy performance metrics computing for

classic search functionality.

Accuracy performance

metrics (for classic search

functionality)

values

Precision

Recall

F1 measure

4/20=0.2=20%

4/4=1

0.33

5.4 Experimental Results and

Discussion

Results: The results indicated that after combining the

proposed SBPR recommender system (as mentioned

in the case study) based on the user’s LinkedIn

business profile (Patty Adel in the case study) with a

user’s inquiry about the process model he wants,

(claim process) the accuracy of the BP model

recommendations can be further improved.

Discussion: As we see in the Table 4 the number

of BP models recommended to user when applying

for our recommender system is decreased to six BP

models while it was at twenty BP models found via

the classic search functionality provided in

repository. The results show (as mentioned in figure

5) that the number of BP models found that fit both

user business profile and manual inquiry of user

(where the user specified the process he wants by

keyword ‘claim’) is decreased by 70 percent

compared to the number of models found via the

classic search functionality. The case study presented

is conducted on our established dataset of 600 BP

models.

6 CONCLUSIONS

In this paper we presented a conceptual framework of

BP model recommendation based on the user social

profile and implemented as a transformation process

in model driven engineering (MDE). The benefit of

such a Framework is to facilitate the process

modeling by reusing existing BP models from the BP

model repository. The recommended BP models are

based on business profile of user. The users’ business

profiles are captured from their online LinkedIn

social network.

We have proposed to integrated our SBPR

recommender to current process model repositories.

without any modifications in repositories’ current

structures. Most of actual repositories offer simple

interfaces, where users can only search by keywords

such as MIT repository or explore the available

content via taxonomical navigation such as in SAP

100

200

300

400

500

600

Our

approach

Classic

Total

models

dataset

600

NumberofBPmodels

numberofBPmodels

0.5

1.5

Precision Recall F1

Accurancy Performance Comparaison

Metrics

Accuracy performance

comparison

Classic SBPR

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments

repository. The performance metrics have shown that

the precision metric in our SBPR recommender

system is more relevant compared to the classic

search functionality. Furthermore, the number of

models recommended is decreased compared to the

number of models found via the classic search

functionality. In this paper also, we have proposed to

implement our SBPR recommender system as a MDE

model transformation process, consequently we have

proposed two metamodels User business profile

metamodel and BP metadata metamodel, to find

correspondences between the two metamodels we

have relied on the mapping specification. Therefore,

we proposed a metamodel for mapping specification

between the two metamodels. We have briefly

presented transformation rules with ATL

transformation language to specify the model

transformation. Once mappings are specified between

the two metamodels, metadata (that reflect business

environment of user) are generated automatically.

these metadata are then used to find the closest

matches in the BP model repository.

We discussed the application of our approach to a

case study where user search for specific BP model

and specify it as a keyword. The results show that the

number of BP models found that fit both user business

profile and manual inquiry of user is decreased by 70

percent compared to the number of models found via

the classical search functionality. The case study is

conducted on a dataset of 600 BP models. In our

future research, we will use a transformation tool to

facilitate the transformation from User profile into BP

model metadata metamodel and then, we will develop

a prototype to validate our approach.

REFERENCES

Ariouat, Hanane, Eric Andonoff, and Chihab Hanachi.

2018. “Process Recommendation Using Context in

Crisis Management: Application to Flood

Management.” In 15th International Conference on E-

Business and Telecommunications (ICETE 2018),

Scitepress, 277–88.

Bézivin, Jean et al. 2003. “First Experiments with the ATL

Model Transformation Language: Transforming XSLT

into XQuery 1.” In Nd OOPSLA Workshop on

Generative Techniques in the Context of Model Driven

Architecture , , 46.

Bobek, Szymon, Mateusz Baran, Krzysztof Kluza, and

Grzegorz J Nalepa. 2013. “Application of Bayesian

Networks to Recommendations in Business Process

Modeling.” In Business Process Modeling In : AIBP@

AI* IA., , 41–50. http://prosecco.agh.edu.pl.

Born, Matthias, Jens Kirchner, and Jörg P Müller. 2009.

“Context-Driven Business Process Modelling.” In The

1st International Workshop on Managing Data with

Mobile Devices (MDMD 2009), Milan, Italy, 6–10.

Deng, Shuiguang et al. 2017. “A Recommendation System

to Facilitate Business Process Modeling.” IEEE

Transactions on Cybernetics 47(6): 1380–94.

Elias, Mturi, and Paul Johannesson. 2012. “A Survey of

Process Model Reuse Repositories.” In , 64–76.

http://link.springer.com/10.1007/978-3-642-29166-

1_6 (September 8, 2019).

Elias, Mturi, Khurram Shahzad, and Paul Johannesson.

2010. “A Business Process Metadata Model for a

Process Model Repository.” In , 287–300.

http://link.springer.com/10.1007/978-3-642-13051-

9_24 (September 8, 2019).

Fellmann, Michael, Agnes Koschmider, and Andreas

Schoknecht. 2014. “Analysis of Business Process

Model Reuse Literature: Are Research Concepts

Empirically Validated?” In Lecture Notes in

Informatics (LNI), Proceedings - Series of the

Gesellschaft Fur Informatik (GI), Gesellschaft fur

Informatik (GI), 185–92.

Gao, Shang, and John Krogstie. 2010. “A Repository

Architecture for Business Process Characterizing

Models.” In , 162–76. http://link.springer.com/10.1007/

978-3-642-16782-9_12 (September 8, 2019).

Gerth, Christian. 2013. 7849 Business Process Models.

Change Management. Berlin, Heidelberg: Springer

Berlin Heidelberg. http://link.springer.com/10.1007/

978-3-642-38604-6.

Hammoudi, Slimane, Wajih Alouini, Denivaldo Lopes, and

Marianne Huchard. 2010. “Towards A Semi-Automatic

Transformation Process in MDA: Architecture,

Methodology and First Experiments.” Journal of

Information System Modeling and Design (IJISMD)

1(4): 48–76. www.igi-global.com.

Hofreiter, Birgit, and Christian Huemer. 2006. “From a

UMM Business Process Model to a Business

Environment Specific EbXML Process.” Journal of

Ecommerce Research 7(3): 138–53.

Hornung, Thomas, Agnes Koschmider, and Andreas

Oberweis. 2009. “A Recommender System for

Business Process Models.” In 17th Annual Workshop

on Information Technologies & Systems (WITS).,.

Imad, Abdullah, Abdullah Elkindy, and M Sc Carl Corea.

2019. “Survey of Business Process Modeling

Recommender Systems.” Koblenz.

Jean-Marie Favre. 2007. “Metamodel Linkedin.”

Wikimedia. https://commons.wikimedia.org/w/index.

php?title=File:Metamodel_Linkedin.jpg&oldid=18006

6553.

Khider, Hadjer, Slimane Hammoudi, Amel Benna, and

Abdelkrim Meziane. 2018. “Social Business Process

Model Recommender: An MDE Approach.” In 2018

Fifth International Conference on Social Networks

Analysis, Management and Security (SNAMS), IEEE,

106–13. https://ieeexplore.ieee.org/document/8554

581/ (September 9, 2019).

Kluza, Krzysztof, Mateusz Baran, Szymon Bobek, and

Grzegorz J Nalepa. 2013. “Overview of

Recommendation Techniques in Business Process

MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development

Modeling.” In In Proceedings of 9th Workshop on

Knowledge Engineering and Software Engineering

(KESE9) , ed. (2013). Overview of recommendation

techniques in business process modeling. In

Proceedings of 9th Workshop on Knowledge

Engineering and Software Engineering (KESE9) (pp. ).

, 46–57.

Koschmider, Agnes, Thomas Hornung, and Andreas

Oberweis. 2011. “Recommendation-Based Editor for

Business Process Modeling.” Data & Knowledge

Engineering 70(6): 483–503.

Laurenzi, Emanuele et al. 2019. “Towards An Assistive and

Pattern Learning-Driven Process Modeling Approach.”

In AAAI Spring Symposium: Combining Machine

Learning with Knowledge Engineering.,.

Leopold, Henrik et al. 2019. “Searching Textual and

Model-Based Process Descriptions Based on a Unified

Data Format.” Software and Systems Modeling 18(2):

1179–94.

Li, Ying et al. 2014. “An Efficient Recommendation

Method for Improving Business Process Modeling.”

IEEE Transactions on Industrial Informatics 10(1):

502–13.

Markovic, Ivan, Alessandro Costa Pereira, and Nenad

Stojanovic. 2008. “A Framework for Querying in

Business Process Modelling.” In Business Process

Modelling. In Multikonferenz Wirtschaftsinformatik, ,

1703–14.

Mendling, Jan, Bart Baesens, Abraham Bernstein, and

Michael Fellmann. 2017. “Challenges of Smart

Business Process Management: An Introduction to the

Special Issue.” Decision Support Systems 100: 1–5.

Zur Muehlen, Michael. 2004. “Workflow-Based Process

Controlling Foundation , Design ,.” In Information

Systems, , 299.

Nadee, Wanvimol. 2016. “Modelling User Profiles For

Recommender Systems.” Queensland University of

Technology.

Rangiha, Mohammad Ehson, Marco Comuzzi, and Bill

Karakostas. 2015. “Role and Task Recommendation

and Social Tagging to Enable Social Business Process

Management.” In Lecture Notes in Business

Information Processing, Springer Verlag, 68–82.

Rosemann, Michael, Jan Recker, and Christian Flender.

2008. 3 International Journal of Business Process

Integration and Management Contextualisation of

Business Processes.

Salman, Aslam. 2019. “Linkedin by the Numbers: Stats,

Demographics, Fun Facts.” https://www.

omnicoreagency.com/linkedin-statistics/ (November

10, 2019).

Schonenberg, Helen, Barbara Weber, Boudewijn Van

Dongen, and Wil Van Der Aalst. 2008. “Supporting

Flexible Processes Through Recommendations Based

on History.” In International Conference on Business

Process Management, Springer, Berlin, Heidelberg.,

51–66.

Setiawan, Mukhammad Andri, Shazia Sadiq, and Ryan

Kirkman. 2011. 87 LNBIP Facilitating Business

Process Improvement through Personalized

Recommendation.

Shahzad, Khurram et al. 2009. “Elicitation of Requirements

for a Business Process Model Repository.” In , 44–55.

http://link.springer.com/10.1007/978-3-642-00328-

8_5 (September 8, 2019).

Shahzad, Khurram, Mturi Elias, and Paul Johannesson.

2010. “Requirements for a Business Process Model

Repository: A Stakeholders’ Perspective.” In Lecture

Notes in Business Information Processing, Springer

Verlag, 158–70.

Weske, Mathias. 2019. “Business Process Management

Architectures.” In Business Process Management:

Concepts, Languages, Architectures, Berlin,

Heidelberg: Springer Berlin Heidelberg, 351–84.

Yan, Zhiqiang, Remco Dijkman, and Paul Grefen. 2017.

“Generating Process Model Collections.” Software and

Systems Modeling 16(4): 979–95.

Yan, Zhiqiang, and Paul Grefen. 2010. LNBIP 66 - A

Framework for Business Process Model Repositories.

Business Process Model Recommendation as a Transformation Process in MDE: Conceptualization and First Experiments