A Model to Specify Inter-company Cooperation

Meriem Kermani

, Mahmoud Boufaida

and Emmanuel Paviot-Adet

LIRE laboratory, Constantine 2 University, Constantine, Algeria

LIP6 laboratoty, Pierre & Marie Curie University, Paris, France

Keywords: Interoperability, Inter-company Cooperation, Learning Agents.

Abstract: The impact of the evolution of the global economic ecosystem, has forced and still force again organizations

to either adopt new behaviour schemas and to deeply change their structure, also to be more open to their

environment as well. These impacts require a new cooperation philosophy from organizations side. Several

studies have focused on the problem of the intercompany cooperation, proposing approaches that provide

interoperability mechanisms. But this remains an open research domain. In this work we will propose a

mediation architecture between different companies. A mediator allows us to create an intercompany

cooperative process, the purpose of this solution is to keep the company architecture and ask the mediator

which is a software-based agent to play an intermediary role between companies, and involve it in making

the transformation between companies. We define a dynamic and cooperative inter-companies model that

combines the agent technology and the decision trees paradigm. This last facilitates making decision by

selecting the services that best meet customer needs, in order to create a composite service.

1 INTRODUCTION

The company environment has strongly evolved in

recent years; it has become more complex and

unpredictable, moreover technological changes have

greatly influenced the organization of the

companies. The integration of these changes

becomes essential to ensure companies’ survival, not

only at a single company, a sector or a given

economy level but at global level (Grefen et al.,

2009) companies are now trying to “squaring the

circle” because they must firstly meet the

competitive challenges by improving industrial

performance in terms of cost, time, etc. Secondly

answer the openness problems. Many companies

have realized the importance of cooperation to

ensure their survival. Cooperative spaces, in which

companies work and react together have emerged in

various forms: virtual company, companies’ network

etc. The establishment and the management of this

type of organization are often based on technical and

information sharing platforms which support and

facilitate the cooperation (Grefen et al., 2009). We

will focus in our work in this kind of relationship

and more particularly to the notion of the on demand

cooperation. We take the definition proposed by

(Boukadi et al., 2010) of the on demand cooperation

as a temporary grouping of partners distributed in

space and time. This grouping is formed from

opportunistic alliances initiated by a company called

the initiator company of cooperation project which

will be dissolved once the opportunity finished.

Such a scenario involves the collaboration of

different parts of a cooperative process that consists

of several processes performed by different partners

in order to meet a common purpose or to seize an

opportunity in the market.

While the information technology innovates

increasingly, but the industrial world is still reluctant

to join dynamic cooperation scenarios, since the

companies’ information system is not suited to this

type of operation.

The service-oriented architecture and Web

service technology seem to provide credible answers

to the needs both internally (information system

support of the cooperation) and externally (services

available to partners). But in the current state,

several obstacles constrain the implementation of the

SOA within the company as well as the forming of

services-based cooperation. It is basically the lack of

implementation method to define the architecture of

services within the company information system. In

addition, there are few works about the construction

methods of the on-demand cooperative process

based on services composition paradigm. To these

318

Kermani M., Boufaida M. and Paviot-Adet E..

A Model to Specify Inter-company Cooperation.

DOI: 10.5220/0004841103180323

In Proceedings of the 10th International Conference on Web Information Systems and Technologies (WEBIST-2014), pages 318-323

ISBN: 978-989-758-023-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

two problems a third one related to the dynamic

nature of the cooperation is added. The analysis of

the problem of services based inter-company

cooperation has led us to the decomposition into

three additional sub issues that shape our research

problem, which are respectively the building

problem of a new framework that ensures the

effectiveness and efficiency of the intercompany

cooperation based on service approach, the problem

of decision making in order to choose the best

services for a given query and the building problem

of the on-demand collaborative process (or the

composition of different services). Our contributions

answer the limitations and problems previously

described by offering mediation architecture

between different companies which are a

cooperation partners, the use of mediator ensures

interoperability between companies and some level

of security. To facilitate decision making and be able

to choose the services that best meet customer needs,

we propose the integration of decision trees in the

software agents of our mediator. The third part of

our contribution is the implementation of

cooperation; this cooperation necessarily passes

through an interconnection of processes of different

companies, using the service composition paradigm.

This paper is structured as following: Section 2

and 3 are consecrated to a study of our proposal

which will be illustrated by an example in section 4.

Section 5 is dedicated to a comparative study

between multiple research works finally in section 6,

we will make conclusions about our work and we

will talk about the extension of it.

2 THE PROPOSED MODEL

After having introduced the concept of Inter-

enterprise cooperation that results in the formation

of business process built from the interconnection of

various processes and the partners have a partial

master on the business process. We propose in this

part mediation architecture between the services of

the companies that are participating to the

cooperation. According to J-L Lascoux (Touzi et al,

2008) the mediator “provides a framework with its

own benchmarks, consisting of operating and

communication rules, and steps process. It begins

with the acknowledgement of the parts position in

terms of legitimacy, to the formalization of an

agreement as satisfactory as possible for the parts”.

In our architecture the mediator is used to perform a

dynamic intercompany process. To ensure the

interoperability of the participating companies we

propose the integration of software agents in our

mediator, learning of these agents is done by

decision trees. The intercompany process (composite

service created) will be published to be used as a

web service, to permit the visibility and to be reused

by other customers. Figure 1 shows an overview of

our architecture.

Figure 1 shows our mediation architecture that

consists of:

 Participating Companies: These companies offer

accessible services via the public part of their

information system. The Information about each

service is stored in a services directory of the

mediator.

 Mediator: is the core of the architecture, it is

acting as an intermediate between the requester

and the partners. It is composed of:

 Interface Agent: it receives and makes the

processing of the applicant request to build a

decision tree, which will be sent to the

coordinator agent.

 Services Directory: at this directory we have

services agents each of these agent is related to a

service proposed by a company. These agents

Figure 1: Architecture for intercompany cooperation.

AModeltoSpecifyInter-companyCooperation

319

represent information about each service, the

information is organized as: a local decision tree

in each agent service, and service address,

communication protocol, and the details of the

servicepurpose.

 Coordination Agent: it allows creating a

composite service to meet the client request. It

has a global decision tree that is created from the

set of local decision trees.

 This tree is used to

make a decision which allows choosing the best

services to meet customer demand.

 Applicant service: it represents the launch

process.

3 LEARNING AGENTS BY

DECISION TREES

To enable the communication of different

heterogeneous companies we propose to combine

the software agents and decision trees.

Figure 2 shows the general structure of the

coordination agent, respecting the three properties of

independence, communication and intelligence. This

architecture is inspired from the modular theory of

(Fodor, 1983) that consists of five modules

managing knowledge, perception, communication,

control and the thinking of the agent.

Figure 2: Components of the Coordinator Agent.



Communication Module is a messages support

between the agent and its coordinator. All

interactions between the coordination agent and

the services agent or the interface agent pass

through this module.

 Mailbox is used to submit messages. This queue

box type FIFO (First in First Out), is used for

storing messages in order to process them

asynchronously.

 Representation module receives the request sent

by the agent service as a tree, and then

reformulates the global tree (the tree that

includes all local trees published by the different

service agents) following the order of criteria

send by interface agent



Information Management Module contains

information related to the service published by

partners (protocol, address, message structure,

description of the roles of services). Each agent

service uses a local ontology to represent

vocabulary and concepts in its application

domain.

 Ontology module deals with the search for

correspondences between agents to perform a

cooperative task.

4 DECISION-MAKING

MECHANISM BASED ON

DECISION TREES

The coordinator agent uses a decision tree to quickly

select the services that best meet customer needs, a

decision tree can be used to clarify and find an

answer to a complex problem. The structure allows

our coordinator agent to take a problem with

multiple possible solutions and display it in a simple

and easily understandable format that shows the

relationship between different events or decisions.

4.1 Structure of the Local Tree

The following figure shows the internal structure of

our local trees:

Figure 3: Internal structure of our local trees.

- Initial Node: each local tree initial node located

in an agent service represents a company service.

-Branches: they connect an interior node”

parent” to a “child” node. They represent the

variable value tested in the parent intermediate node

to the child node. Each branch corresponding to a

question, allows choosing the path to follow.

Rule 1: the path to go in the tree: if the answer is

yes, then the left path will be taken, otherwise it

must move to the right one.

-Intermediate Node: represents the deals offered

WEBIST2014-InternationalConferenceonWebInformationSystemsandTechnologies

320

by each service, each intermediate node

corresponds to a response to determine the deal.

Each answer then determines the next question.

Rule 2: each non-terminal node corresponds to a

test P(x) which evaluates the predicate P in Example

-Leaf: each leaf refers to a class.

4.2 Training Algorithms by Decision

Tree

The main training algorithms by decision trees are

CART (Breiman et al, 1994) and C4.5 (Quinlan,

1984). We present below some of their

characteristics. The first step is to build a small tree

which consists of the most of data. The idea is to

divide recursively and efficiently as possible the

training sets examples by defined tests using

attributes until we get subsets of examples which

contain only examples belonging to the same class.

This idea leads to a Top-Down construction methods

that’s mean constructing the tree from the root to the

recursive leaves. We can then define a general

scheme of algorithm.

Generic Training Algorithm:

5 EXAMPLE: EMERGENCY

CASE

We will treat in this part the realization of a response

process of an emergency case which is an accident,

using three existing entities named: Hospital, SAMU

and Police.

 The Hospital service: each hospital propose

a description including its position, in our case we

will need a GPS coordinates of the hospital.

 The SAMU service: provides a description

of its services such as the emergency care team, the

possible route and the expected duration of each

route.

The Police service: provides the position of the

police stations and available patrols.

 The description of each hospital the SAMU

and The Police services is published in an agent

service and represented by a local decision tree, and

the coordination agents representing the global tree

for each existing entities.

5.1 Operational Scenario

Using the three services mentioned previously, we

will elaborate an emergency response scenario to

carry out a sequence of actions for each accident

report. We wish to create here a response emergency

service which is called 333. This service will allow

users to report an accident at a given position. The

service will take the necessary steps and return to the

user the expected intervention delay in seconds. The

internal scenario which will lead the interactions

between services is the following:

The coordinator agent will firstly ask the

accident base to determine whether the reported

accident is known, if so a negative value is simply

returned to the service requester. Otherwise, we will

first find the nearest hospital to the accident position,

we use GPS data of the hospital and we determine

both the SAMU and the police to be sent. We then

add this to the accidents base before returning to the

user the intervention time that was given by the

SAMU.

5.2 Creation of Trees

We have chosen to use the SIPINA method for the

generalization of our local trees, SIPINA is software

but it is also a training method. It generalizes the

trees by introducing an additional operation which is

the fusion, during the induction of the prediction

model. The figure 4 shows the generalization of the

tree from Excel files, after following all method

steps we obtain the following tree. After creating all

local trees, the coordinator agent generates a global

tree that includes all local trees; this tree that has

been created manually, until the definition of an

algorithm which allows the fusion of the local trees

according to a test on each arc in order to create a

tree that includes other trees by category.

AModeltoSpecifyInter-companyCooperation

321

Figure 4: Generalization of the tree from Excel using

Sipina Version 3.2.

5.3 Some Implementation Aspects

Figure 5: Communication between agents JADE.

Figure 5. shows the structure of our agents, and

some messages exchanged between its, using the

Jade platform to create the behavior of agents and to

examine the feasibility of our proposition. For

example coordinator agent of SAMU service named

"CA-SAMU" sends a request to the service agents

"SA-SAMU" the representatives of the different

SAMU, then this services respond with a messages

"inform" given their position, the coordinator agent

selected the service that best meets the needs and

sends a message "confirm" to the service agent

concerned, the traces of the messages exchanged

between agents are made with the help of sniffer

agent. The third part of the figure shows the

structure of a message "request" sent by the agent

"CA-SAMU" to the "CA-SAMU" agent.

6 SOME RELATED WORK

Several models have been proposed to establish the

enterprise collaboration. The proposed works have

the information system at the center of their

concerns and qualify it as an indispensable support

for the implementation of the company strategy. The

authors of some research work (Indulska et al, 2006)

(Henderson et al, 2007) (Rolland et al, 2009) were

attached to the analysis and design domain of

process-oriented information systems. Several

methods and tools of information systems

engineering have emerged such as component-based

approaches, MDA (Model Driven Engineering).

These methods have been proposed to meet the

evolution of the current information system context:

evolution in terms of the information system

architecture (heterogeneous and distributed),

evolution in its use (open and accessible for different

users). Since a few years the concept of Service

Oriented architecture (SOA: Service Oriented

Architecture) has been rapidly spread and widely

accepted as a supporting architecture of the company

information system. We studied the overall work

according to the approaches and technologies used

the nature of collaborative processes if it is static or

dynamic, and the field possible for the use of the

proposed solution.

(Touzi et al, 2008) adopt a non standardization

approach, which proposes the design of a mediation

system between information systems. The proposed

practices are similar to the MDE and allow one to

anticipate the tooling of a model transformation

workshop. This method was also adopted in (Truptil

et al, 2011), to solve a crisis management problem,

the approach used in (Hui et al, 2010) is a

standardization approach that consists on the

proposition of a framework where all the

participating company must respect it.

Among the works that have used the agent

paradigm for solving the inter-company cooperation

problems, the authors, (Brahimi et al, 2009) (Namin

et al, 2006) have proposed global architectures,

integrating the technology of web services. The

difference that (Namin et al, 2006) proposed the

integration of web services and software agents

within the internal structure of the company and by

adopting software agents within the UDDI registry;

from the company side. Few studies have treated the

notion of dynamic cooperation. For instance, the

work of (Boukadi et al, 2010) offers such a

mechanism of cooperation. (Mallek et al, 2011) have

a different vision. They proposed an approach to

detect problems before a real collaboration, then

WEBIST2014-InternationalConferenceonWebInformationSystemsandTechnologies

322

analyzing and finding solutions to each partner. A

few studies cited above that have used software

agents and they have not addressed the decision-

making of these agents. In this work we propose a

distributed system, and when it comes to designing

this type of system, agent technology is suitable,

because MAS not only allow the sharing or

distribution of knowledge, but also the achievement

of a common goal. Agents in our work learn the

decision from the user in certain situations to

support him in other situations. The decision

problem in the case of the cooperative distributed

systems is a complex problem. Then, the training

algorithm of the agent must meet several

requirements: interactive, explicative, online, not

handling only the quantitative but also the

qualitative parameters. Also we must divide our

decision problem into several problems. In our case,

we choose to use decision trees, because they can

model simply graphically and quickly a complex

measured phenomenon. The aim of the use of

training agents with the paradigm of decision trees is

to select the best services that meet customer needs.

7 CONCLUSIONS

In this paper, we have proposed an architecture to

create a dynamic process of cooperation; our

architecture can be used in the case of B2B

(business-to-business), B2C (Business-to-Consumer)

or B2A (business-to- administration). This solution

is based on approach of no-standardization;

companies keep their structure and offer services.

The mediator is involved in making the

transformation between the services to create an

intercompany collaborative process. We have

integrated decision trees in a module of agents called

representation module. The use of decision trees can

reduce the response time to choose the best service.

In order to better improve our approach, we will use

the reinforcement learning (RL) algorithm, which

has long been promising methods for enabling an

autonomous agent to improve its behavior on

sequential decision-making tasks. The next step is to

specify a transformation model for integrating the

decision trees in the agent modules.

REFERENCES

Boukadi, K., Vincent, L., Ghedira, C., Maamar, Z., 2010.

CSMA: Context-Based, Service-Oriented Modeling

and Analysis Method for Modern Enterprise

Applications. In International Journal of

Organizational & collective Intelligence, pp 1-28.

Brahimi, M., Boufaida, M., Seinturier, L., 2009 Multi-

Agent Architecture for Developing Cooperative E-

Business Applications. In International Journal of

Information Systems and Supply Chain Management

(IJISSCM).

Breiman, L., Friedman, J. H., Olshen, R. A., Stone, C.,

1984. Classification and Regression Trees,

Wadsworth. Inc., Belmont, California.

Fodor, J., 1983. The Modularity of the Mind, Cambridge,

Mass, MIT Press.

Grefen, P., Kouvas, N., Weichhart, G., Eshuis, R.,

2009. Dynamic business network process management

in instant virtual enterprises, Computers inIndustry,

pp. 86–103.

Henderson, S., Gonzalez, C-P., Ralyté, J., 2007 Situational

method engineering: chunks or fragments. In the

CAiSE ForumTrondheim, pp 89-92.

Hui, L., Anne, F. Cutting, D., Bourey, J.P., 2010. Use of

Ontology for Solving Interoperability Problems

between Enterprises. In the 11th IFIP Working

Conference on virtual enterprises.

Indulska, M., Green, P., Recker, J., Rosemann, M., 2006.

Business Process Modeling: Perceived Benefits. In the

28th International Conference on Conceptual

Modeling Springer, Gramado, Brazil .

Mallek, S., Daclin, N., Chapurlat, N., 2011. An Approach

for Interoperability Requirements Specification and

Verification. In Third International IFIP Working

Conference, pp 89-102.

Namin, A-S., Shen, W., Ghenniwa, H., 2006.

Implementing enterprise collaboration using web

services and software agents. In International Journal

of Nerworking and Virtual Organisations, pp 185-201.

Quinlan, J. R., 1984. C4.5: Programs for Machine

Learning, Morgan Kaufmann.

Rolland, C., 2009. Method engineering: towards methods

as services, Software Process: Improvement and

Practice, 14(3), pp. 143-164.

Touzi, J., Bénaben, F., Pingaud, H ., 2008. Prototype to

Support Morphism between BPMN Collaborative

Process Model and Collaborative SOA Architecture

Model. In the 4th International Conference on

Interoperability for Enterprise Software and

Applications, pp 145-157.

Truptil, S., Bénaben, F., Pingaud., P., 2011. On-the-fly

adaptation of crisis response information system. In

Conference on Management of Emergent Digital

EcoSystems. pp114- 121.

AModeltoSpecifyInter-companyCooperation

323