A SIMULATION-BASED DIFFERENCE DETECTION

TECHNIQUE FOR BOTTOM-UP PROCESS RECONCILIATION

Xi Chen and Paul W. H. Chung

Department of Computer Science, Loughborough University, Loughborough, Leicestershire, United Kingdom LE11 3TU

Keywords: Business process, reconciliation, simulation, B2B e-commerce.

Abstract: With the increasing dynamic and changing business environment, bottom-up approaches for business

process collaboration is currently receiving a great deal of attention in the research community. Bottom-up

approaches are seen to be more flexible then top-down approaches. However, none of the available

techniques for process collaboration are suitable for process reconciliation, which is a common problem

when different organisations have to work together. In order to address the issue in a bottom-up way, a

simulation-based technique for detecting differences between any two given processes is proposed. It is

based on the extended definitions of process compatibility for collaboration and is the core of the process

reconciliation mechanism.

1 INTRODUCTION

In modern enterprises, workflow technology is

commonly used for business process automation.

Established business processes represent successful

work practice and become a crucial part of corporate

assets. In the era of Internet, business processes are

unlikely to remain within the boundary of a single

organisation. In response to the need for B2B e-

commerce, the concept of business process

collaboration is emerging.

However, before any two business partners can

proceed in conducting B2B e-commerce

transactions, their business processes that are

involved in the transactions must be compatible with

each other at business level (Yang and Papazoglou,

2000), i.e. they have a commonly agreed sequence

of exchanging collaborative messages (e.g. a

business object like a purchase order or a service

invocation request). In general, there are two general

approaches to achieve compatibility for process

collaboration between two trading partners, namely,

top-down and bottom-up. A top-down approach

normally involves that people meet and discuss the

collaboration, design the collaborative process and

implement it locally. On the contrary, a bottom-up

approach derives collaborative process from local

processes, which is known a difficult task. However,

because the top-down approach is labour intensive

and expensive (Wombacher, 2005), it is necessary

and worthwhile to explore the feasibility of the

bottom-up approach in the face of an increasingly

dynamic and changing business environment.

Current techniques for process collaboration are

not able to provide sufficient computer assistance for

bottom-up process reconciliation for a number of

reasons. First, the definition of absolute

compatibility, adopted by the top-down approach, is

too limited. Other categories of compatibility will

need to be identified. Secondly, a process

reconciliation mechanism is required to consider all

the relevant activities. Thirdly, as the core of the

mechanism, a technique for process difference

detection is needed to be able to address the

differences encountered and guide the user towards a

possible common process.

In section 2, definitions of process compatibility

for collaboration are reviewed and new definitions

proposed. On these definitions process reconciliation

activities are based. Section 3 depicts the desired

process reconciliation mechanism with emphasis on

support for a unilateral decision-making process. In

section 4, a simulation-based technique for detecting

process differences is proposed and is explained by

walking through an example. Conclusion is drawn in

chapter 5 and future work is described.

The activity-based workflow modelling

formalism (Bi and Zhao, 2004) is used in the rest of

the paper as it is useful visual representation of

Chen X. and W. H. Chung P. (2007).

A SIMULATION-BASED DIFFERENCE DETECTION TECHNIQUE FOR BOTTOM-UP PROCESS RECONCILIATION.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - SAIC, pages 72-77

DOI: 10.5220/0002394600720077

 SciTePress

business processes. In this representation a vertical

synchronisation bar is used as the symbol for an

AND vertex; a circle with a cross inside is for an

XOR vertex; a rectangle is for a normal activity

vertex and an arrow is a directed arc. The only

difference from the formalism is the type of split or

join (AND or XOR) of a routing vertex is expressed

by the pre-condition or post-condition of its

neighbour activity vertex, which makes the routing

vertex merged with its neighbour. Further more,

when needed, an activity vertex explicitly shows its

role in the collaboration as either a message sender

(s) or receiver (r) (Chen and Chung, 2006) as a

superscript.

2 PROCESS COMPATIBILITY

FOR COLLABORATION

The purpose of reviewing the definitions of process

compatibility for collaboration is to clarify the goal

that process reconciliation needs to achieve if a

bottom-up approach is followed. According to

Hiltrop and Udall (1995), one of the essential

principles of negotiation is to get what both sides

want rather than to win at any cost. Apart from the

unanimously agreed absolute compatibility, another

two types of compatibility can be named as

deadlock-free compatibility and reconcilable

compatibility (Wombacher, 2005; Krukkert, 2003).

These three types of compatibility for collaboration

are defined below.

Definition 1: Absolute compatibility. Two

abstract collaborative processes have the same set of

activity vertices, routing vertices and arcs.

Definition 2: Deadlock-free compatibility. If the

difference between two processes are only XOR

activities on the receiving process and the

corresponding sending activities do not split into

XOR branches then the two processes are deadlock-

free compatible. For example, in Figure 1, process A

and B are different but are deadlock-free compatible.

When B informs A the only available option by

sending a message from B.b

to A.b

, a deadlock

situation will not arise. Therefore, no adjustment is

required for both sides.

Figure 1: Deadlock-free compatibility.

Definition 3: Reconcilable compatibility. Two

processes may appear differently but are

reconcilably compatible if they have the same set of

activity vertices and the maximum intersection of

the sets of possible paths contains at least one path

that leads to success. A path denotes a possible

execution sequence of all the activities that can be

reached based on the current process definition.

As illustrated in Figure 2, process A is different

from process B but there exists process C that can be

successfully traced through both A and B. Thus,

process C can be adopted, which meets the

requirements of A and B, and collaboration can

proceed.

Figure 2: Reconcilable compatibility.

These definitions collectively form a set of

acceptance criteria when considering in a bottom-up

manner whether any two given processes can be

simply adjusted to be compatible for collaboration.

For processes that are not compatible, i.e. they

do not meet any one of above definitions, then

differences between them must be detected and

appropriate changes must be made by either or both

partners in order to bring about collaboration.

3 PROCESS RECONCILIATION

MECHANISM

Bilateral negotiation is an effective way of

reconciling differences in a distributed manner (Li et

al., 2003). According to Li et al., such a negotiation

comprises a series of unilateral decisions within the

control of an underlying negotiation protocol. Since

it is common that more than one discrepancy exist

between two processes, partners involved are very

likely to negotiate and make decisions on multiple

issues, which makes the bilateral negotiation a

multi-attribute (or multi-issue) one (Fershtman,

1990). Whether to apply a simultaneous or

A SIMULATION-BASED DIFFERENCE DETECTION TECHNIQUE FOR BOTTOM-UP PROCESS

RECONCILIATION

sequential protocol for a multi-attribute negotiation

depends on the problem itself. This is because

discrepancies between processes are often

interdependent, in order to prevent simultaneous

controversial decisions from being made, the

negotiation protocol can only be set as a sequential

one in the form of alternate proposal of counteroffer

after the initial offer. Therefore, within a bilateral

negotiation process are many unilateral decision-

making steps that take place on both sides.

In the light of the requirement of privacy and the

complexity of the decision making process, it is

assumed that within a distributed B2B environment,

the unilateral decisions are made by people on both

sides of the negotiation as shown in Figure 3. This

process is repeated until a modified collaborative

process is completely formed or collaboration is

abandoned.

Figure 3: Unilateral decision-making process.

Most current techniques for process matching

take a static view towards the differences between

two processes. Few have considered process

reconciliation when mismatches are encountered. If

these techniques were applied directly to process

reconciliation, they either confine themselves to

minor passive adaptations (Krukkert, 2003; Du et

al., 2005; Wombacher, 2005) or handle drastic

changes without giving user any selection right

(Yeoh et al., 2004). Juan (2006) proposes a string

comparison approach to tackling process logic

differences however the preliminary case study

shows that the approach is restrictive because it

requires to express the process in digraph form into

strings that is on sequential level. Demanded by the

interactive and repetitive nature of the process

reconciliation task, discrepancies must be identified

according to the progress of reconciliation. Also, the

current reference process needs to be continuously

updated to reflect the user’s decision regarding the

previous discrepancy. Such dynamism can also be

seen from Figure 3.

4 SIMULATION-BASED

PROCESS DIFFERENCE

DETECTION TECHNIQUE

Since the unilateral decision-making is based on the

discrepancy currently identified, it is required to

detect and prompt the discrepancy to the user

appropriately. Focusing on the process diagram or

the corresponding adjacent matrix would not

contribute further to the desired manner of

difference detection. What these techniques can

reveal are merely structural differences between two

digraphs. From process logic’s point of view, such

differences are trivial. Therefore, the desired

technique should be able to reveal process logic

information.

In theoretical computer science, a simulation

pre-order describes a relation between two state

transition systems that one system behaves in the

same way as the other or one simulates the other.

Although being used only to match two

processes by identifying whether any possible

common paths exist, the simulation-based technique

proposed by Krukkert (2003) does suggest another

view on the problem of process difference detection.

According to Krukkert, an activity-based process

diagram can be converted to a state transition system

(STS) if several prerequisites are met. Related

conversion algorithms are also provided.

Therefore, on the one hand, a simulation-based

technique can be used to identify common paths that

exist between two processes if there are any, which

meets the need of matching for reconcilable

compatibility. On the other hand, even if no such

common path exists, it can be used to reveal to what

extend common states exist as well as from which

point difference occurs. The differences encountered

in the way can be further compared, analysed and

prompted to the user to support the decision-making

task. Since the simulation-based technique is only

valid for processes with the same number of

vertices, a pre-treatment and a post-treatment are

required to deal with unmatched vertices. The

algorithm is constructed as in Table 1.

ICEIS 2007 - International Conference on Enterprise Information Systems

Table 1: Simulation-based difference detection algorithm.

Pre-treatment

Conduct bipartite matching of vertices of both process diagrams, extract only

the matched vertices, record the unmatched vertices and their causal relations

with the matched ones of both sides;

Check for isomorphic sub-structure and replace them with single dummy vertex;

Core part of the difference detection algorithm

WHILE maxCommonPathFlag == FALSE AND terminationFlag == FALSE

STSD_A = activityDtoSTSD(activityD_currRefA);

STSD_B = activityDtoSTSD(activityD_currRefB);

FOR (currLayer = 1; currLayer <= matchedVertexNum; currLayer++)

commonState = StateComparison(currLayer);

IF(commonState)

maxCommonPath = maxCommonPath + commonState;

verifyMaxCommonPath(maxCommonPath);

IF(currLayer == matchedVertexNum)

maxCommonPathFlag = TRUE;

break;

ELSE

dispFwd = getDispFwd();

dispBkwd = getDispBkwd();

adjustmentSuggestion = getAdjustmentSuggestion(dispFwd, dispBkwd);

tempActivityD = adjust(adjustmentSuggestion, activityD_currRefA);

association = checkAssociation(adjustmentSuggestion);

IF(association)

promptUser(theUnmatchedVertices);

decision_unmatched = getUserDecision();

recordDecision(decision_unmatched);

promptUser(adjustmentSuggestion, tempActivityD);

decision = getUserDecision();

adjust(decision, activityD_compromising);

END-FOR

END-WHILE

Post-treatment

adjust(decision_unmatched, activityD_compromising);

Prompt the remaining unmatched vertices and unique exclusive OR branches (if

any) as discrepancies to the user;

decision_unmatched = getUserDecision();

adjust(decision_unmatched, activityD_compromising);

Figure 4: Example processes A, B.

The example illustrated in Figure 4 is used to

explain the algorithm.

After a bipartite matching, a, c, d, e, f and g are

identified as matching vertices. ‘b’ is uniquely

possessed by A whilst ‘m’ and ‘n’ by B. No

isomorphic sub-structure is identified. After the

matching vertices being deposited, process A

and

are the results (illustrated in Figure 5), which are

converted to state transition system graphs as shown

in Figure 6.

A SIMULATION-BASED DIFFERENCE DETECTION TECHNIQUE FOR BOTTOM-UP PROCESS

RECONCILIATION

Figure 5: Result of matched vertices’ extraction.

Figure 6: 1st round comparison between STS graphs of A

and B

Common states are examined in a forward

direction. When discrepancy is encountered, it is

recorded as forward discrepancy (dispFwd) and a

backward examination is carried out with the

encountered backward discrepancy (dispBkwd)

recorded. Since the vertices involved in A

.dispFwd

and A

.dispBkwd are not the same, no swap

operation is required. The A

.dispFwd is by default

selected as the current adjustment suggestion and is

prompted to the user as “‘ce’ should be moved

immediately in front of ‘f’ in process A

” together

with a corresponding activity diagram

representation.

Figure 7: Activity diagrams of A

and B

Assuming that user A accepts this suggestion

and therefore process A

needs to be adjusted

accordingly to A

whilst B

remains the same as B

whose activity diagrams and corresponding STS

graphs are shown in Figure 7 and Figure 8.

Figure 8: 2nd round comparison between STS graphs of

and B

During the second round comparison, four more

common states are identified, which are ‘a | ce’, ‘ac |

e’, ‘ae | c’ and ‘ace | f’. Following on from this,

.dispFwd and A

.dispBkwd are identified and the

involving vertices are evaluated as the same, which

implies the adjustment suggestion should be a

vertices swap operation between ‘g’ and ‘d’ in

process A

. After being prompted both in words and

graphically, assuming that user A rejects the

discrepancy this time, process B

is expected to

make concession by being adjusted instead, i.e.

swapping ‘g’ and ‘d’ in B

to form B

whilst, A

remains the same as A

. Shown in Figure 9 are the

resulting activity diagrams of A

and B

. After the

third round comparison between STS graphs, a

common path is found, whose corresponding activity

diagram C is shown in Figure 10.

Figure 9: Activity diagrams of A

and B

Figure 10: Common path identified between A

and B

ICEIS 2007 - International Conference on Enterprise Information Systems

Vertices that have no match (b, m and n) are

highlighted to user A for further decisions. C

shown in Figure 11 as one of the possible resulting

common collaborative process following

organisation A’s unilateral decisions.

Figure 11: One possible resulting common collaborative

process following organisation A’s unilateral decisions.

When process C

, as the counteroffer, is passed

to organisation B, the same procedure is followed by

user B to carry out their own decision-making. Other

issues, e.g. strategy of unilateral decision-making,

negotiation termination condition, are also important

but beyond the scope of this paper.

5 CONCLUSIONS

The simulation-based process difference detection

technique is proposed to provide support during

distributed process compatibility negotiation by

helping users concentrate on a series of adjustment

suggestions to agree on a common path as soon as

possible. With the support of such a technique,

human efforts are saved from the labour-intensive

task and corporate assets in terms of business

processes are preserved and put into good use. Also,

the ability to start from two predefined process

logics enables the technique to be used from bottom

up, which makes it possible to replace the expensive

top-down approach to cross-organisational process

reconciliation. Furthermore, the technique can also

be applied in the area of process compliance

(Cheung, 2003) as well as process benchmarking

(Juan and Ou-Yang, 2005; Juan, 2006), in which

customer defined processes are checked for

compliance issues against certain standard or best-

of-breed process.

In addition to preliminary case studies having

been carried out, the effectiveness of the technique

needs to be further evaluated through a full range of

real life business processes. It is also envisioned that

a comprehensive business process collaboration

framework is needed to take full advantage of such a

technique, within which the execution components

are mentioned in Chen and Chung (2006).

REFERENCES

Bi, H., Zhao, J., 2004. Applying Propositional Logic to

Workflow Verification. Information Technology and

Management 5, 293–318, 2004.

Chen, X., Chung, P., 2006. Cross-Organisational

Workflow Enactment Via Progressive Linking by

Run-Time Agents. In Ali, M. and Dapoigny, R. (Eds.):

Proceedings of 19th International Conference on

Industrial Engineering and Other Applications of

Applied Intelligent Systems, IEA/AIE 2006, LNAI

4031, pages 54-59, 2006.

Cheung, Y.C., 2003. COMPLIANCE FLOW - AN

INTELLIGENT WORKFLOW MANAGEMENT

SYSTEM TO SUPPORT ENGINEERING

PROCESSES, Ph.D. Thesis, Loughborough

University, 2003.

Du, Z., Huai, J., Liu, Y., Hu, C., and Lei., 2005. IPR:

Automated Interaction Process Reconciliation. In

Proceedings of IEEE/ACM International Conference

on Web Intelligence (WI), 2005.

Fershman, C. The importance of the agenda in bargaining.

Games and Economic Behavior, 2(224-238), 1990.

Hiltrop, J., Udall, S., 1995. The Essence of Negotiation,

Prentice Hall, London 1995.

Juan, Y.C. and Ou-Yang, C., 2005. A process logic

comparison approach to support business process

benchmarking, International Journal of Advanced

Manufacturing Technology, Vol. 26, pages 191-210,

2005.

Juan, Y.C., 2006. A String Comparison Approach to

Process Logic Differences between Business Process

Models. In Proceedings of the Joint Conference on

Information Sciences 2006, Atlantis Press, October

2006. doi:10.2991/jcis.2006.23

Krukkert, D., 2003. Matchmaking of ebXML business

processes, Technical Report IST-28584-

OX_D2.3_v.2.0, openXchange Project, Oct 2003.

Li, C., Giampapa, J.A., and Sycara, K., 2003. A Review of

Research Literature on Bilateral Negotiations, Tech.

Report CMU-RI-TR-03-41, Robotics Institute,

Carnegie Mellon University, November, 2003.

Wombacher, A., 2005. Decentralized establishment of

consistent, multi-lateral collaborations, PhD Thesis at

Technical University Darmstadt, Facultiy of

Informatics, 2005.

Yang, J., Papazoglou, M., 2000. Interoperation Support for

Electronic Business. COMMUNICATIONS OF THE

ACM June Vol. 43, No. 6, 39-47, 2000.

Yeoh, M.L., Chung, P.W.H., Anumba, C.J., El-Hamalawi,

A., Motawa, I.A., 2004. Process change identification

using workflow specification matching. In

Proceedings of the Tenth Americas Conference on

Information Systems, New York, New York, August

2004.

A SIMULATION-BASED DIFFERENCE DETECTION TECHNIQUE FOR BOTTOM-UP PROCESS

RECONCILIATION