J. Software Engineering & Applications, 2008, 1: 83-87
Published Online December 2008 in SciRes (www.SciRP.org/journal/jsea)
Workflow Mining of More Perspectives of Workflow
Peng Liu
1
, Bosheng Zhou
2
1
School of Computer Science and Technology, Beijing University of Aeronautics and Astronautics, Beijing 100191, China,
2
School
of Computer Science and Technology, Beijing University of Aeronautics and Astronautics, Beijing 100191, China
Received November 27
th
, 2008; revised November 30
th
, 2008; accepted December 1
st
, 2008.
ABSTRACT
The goal of workflow mining is to obtain objective and valuable information from event logs .The research of workflow
mining is of great significance for deploying new business process as well as analyzing and improving the already
deployed ones. Many information systems log event data about executed tasks. Workflow mining is concerned with the
derivation of a graphical process model out of this data. Currently, workflow mining research is narrowly focused on
the rediscovery of control flow models. In this paper, we present workflow mining of more perspectives of workflow to
broaden the scope of workflow mining. The mining model is described with GBMS’s VPML and we present the entire
model’s workflow mining with the GBMS’s VPML.
Keywords: Workflow Mining, GBMs, VPML
1. Introduction
Workflow technology continues to be subjected to
on-going development in its traditional application areas
of business process modeling and business process
coordination, and now in emergent areas of component
frameworks and inter-workflow, business-to-business
interaction. Addressing this broad and rather ambitious
reach, a large number of workflow products are
commercially available, which see a large variety of
languages and concepts based on different paradigms.
In 1993, the standardization organization of workflow
technology—Workflow Management Coalition (WFMC)
was built. The definition of workflow in WFMC is as
follows: workflow is concerned with the automation of
procedures where documents, information or tasks are
passed between participants according to a defined set of
rules to achieve, or contribute to an overall business goal
[1]. So, all kinds of activities of enterprises are organized
and corresponded by business processes modeling and
defined business logic relation. Workflow model is the
process model that can be executed in Computer and its
performance can be analyzed with the executing result. In
selection of workflow engine, it must review the analysis
ability of the process model.
Workflow mining means the knowledge discovery of
workflow system, which can be induced by the definition
of data mining. The ultimate target of workflow mining is
to mine the transactional logs of workflow system, and to
discover the knowledge of workflow including workflow
models mining, workflow performance mining and
workflow models improving.
Most workflow mining research is aimed at the
rediscovery of explicit control flow models, which are
used to specify the behavior of a process. We believe that
this approach limits the scope and utility of workflow
mining by neglecting the important notion that workflow
is much more than control flow. In fact, the behavior of a
process is but one of its perspectives. In [2], Wang Lei
and Zhou Bosheng present four major perspectives
(Behavior Model, Information Model, Resource Model,
and Coordination Model) of a process model that have
emerged from the disciplines that helped shape the
workflow area.
2. The Research Basis
2.1 GBMS and VPML
VPML [3] (Visual Process Modeling Language) is a
graphic language that supports a special process
definition. Process is a set of transformations of input
elements into products with specific properties,
characterized by transformation parameters; VPML
describes the process with visual process diagram and
relevant text specification. The diagram shows the
structure of the process and the specification shows the
attribute of all items in the diagram. The process has a
higher degree in visualization and formalization. The
process model built in VPML can simulate. It was proved
that VPML not only describes a whole complete process
model, but also is a process modeling language with rich
functions, visual diagram, easy learning, flexible
application.
GBMS [4] (Government Business Modeling System)is
a modeling system oriented E-government. GBMS not
only supports the Government business’s process
hackling, modeling, simulating and optimal restructure

84 Workflow Mining of More Perspectives of Workflow
that implements interdepartmental business integration
and the sharing of information resources, but also organic
integrates the government business modeling with the
requirement extraction and analysis of the E-government
Business Application System. GBMS completely describes
the business model in the view aspect of process,
organization, resource, information and collaboration and
is process-centered, organic integrates five views and
keeps each view’s consistency and integrity. GBMS lays
the foundation for design and implementation of the
Business Application System in E-government area.
2.2 The Background and the System Architecture
The research background of this paper is that The
Construction of Oriented E-government Software
Component Library. First, using GBMS, we build
process model, organization model, information model,
collaboration model, resource model and behavior model.
The process model is the core of these models, which
clearly describes the government business and finds out
the shared resources. Meanwhile by accumulating plenty
of the government business models, it not only gradually
establishes E-government business pattern base, designs
and builds component library oriented E-government and
sharing in departments, but also unifies the standard of
the government business process and avoids the duplicate
investment of E-government. On the base of the
component library, it executes the models on jBPM’s
workflow engine and takes the rapid response to the new
requirement in the E-government area, assembles the
components of the E-government component library, and
rapidly builds relevant the E-government business
application system. It can build the system on demand
and reduce production costs.
The system’s architecture is given in Figure 1. The
whole project has 5 parts:
(1) GBMS; (2) Component System; (3) jBPM
Workflow Engine System; (4) Automatically Generating
System; (5) Workflow mining System.
Figure 1. The system architecture
This paper extends the paper “The Research on the
Modeling Transformation From GBMS to jBPM” [5], and
presents how to mine the entire model from the
E-government business system in order to validate the
business model built in GBMS.
2.3 Semantics of GBMS and VPML
DEFINITION 1. GBMS as a 6-tuple
PM={P, A, R, Control, Support, Input, Output} where
P={Product
1
,Product
2
,…Product
j
} is a set of products;
A={a
1
,a
2
,…a
m
} is a set of Activities; R={r
1
,r
2
,…r
k
} is a
set of resources; Control is a set of relations of
controlling; Support is the relations of resources
supporting activities, Support
⊆
AxR; Input is the
relations of activities and input products，Input
⊆
AxP;
Output the relations of activities and output products，
Output
⊆
AxP.
We need to define some assistant objects based on the
above definition.
State (i): i ∈ P ∪ R, represents the states of element i. It
is enumeration type: enum={able,disabled} means the
states can be either able or disabled.
Status (a): where a ∈ A, represents the status of the
activity a. In GBMS Model, activity has 2 statuses,
START and END.
Source(c): where c ∈ Input ∪ Output, means the set of
source objects related by c.
Target(c): where c ∈ Input ∪ Output, is a set of target
object related by c.
Prod_Source(a):where a ∈ A, represents the set of
input products related with activity a.
Prod_Target(a):where a ∈ A, means the set of all
output products related with activity a
Resource(a): where a ∈ A, the set of all resources
related with activity a.
Role(a): where a ∈ A, represents the set of all roles
related with activity a.
DEFINITION 2. The GBMS Data Model is used to
describe the information perspective of workflow. GBMS
Data Model is the 4-tuple(P, A, Input, Output).
Input(p, a):is that the Activity a need Product p to run.
Output(p, a) is that Activity a is running to produce the
Product p.
DEFINITION 3. The GBMS Organization Model is
used to describe the organization perspective of workflow.
GBMS Organization Model is the triple (A, R, Support):
A is the set of GBMS’s Activities; R is the set of
resources in GBMS; Support is the relations of resources
supporting activities;
Support (a, r) is that the Resource r supports the
Activity a to run.
DEFINITION 4. S is a subset of P, which represents the
set of all source products.

Workflow Mining of More Perspectives of Workflow 85
{|Input Output,
( )}
SS P c
Source cS
=⊂∀ ∈∪
∩= Φ
DEFINITION 5. E is a subset of P, which represents the
set of all the final products.
{|Input Output,
arg( )}
EE P c
Tet cE
=⊂∀∈∪
∩= Φ
DEFINITION 6. O(a) represents the constraints.
Or
( ){,Pr_( ),
Re( ),( )|
()&&()
&&( )}
O aaApodSource a
rsource auRole a
StatepableState rable
State uable
= ∈∈
∈ ∈
== ==
==
3. The Workflow Mining Algorithm
It is theoretically proved that VPML is equivalent
Petri-Net [6]. So this paper uses the mining algorithm
based on the Aalst’s α-algorithm [7,8,9].
(1) Constructing dependence frequency table (D/F-
table);
By given activities a and b, I) activity a and activity b’s
appearance frequency #a and #b; II) b is directly ahead of
a : #b<#a; III) b directly succeeds a : #a>#b; IV) b is
ahead of a: #b<<<#a; V) a is ahead of b: #a>>>#b; VI)
the degree of dependence between a and b: #a#b;
(2) Mining activities relation table (R-table) by
D/F-table;
Based on the D/F-table, mining the basic activities
relations (a>
w
b ,a
w
b, a#
w
b, a//
w
b) [8];
(3) Reconstructing the workflow net by R-table and
α-algorithm.
4. More Perspectives of Workflow Mining
In this section, we present a sample of more perspectives
of Workflow Mining. The workflow event log is shown
in Table 1. First, we do some reasonable assumptions
because of the workflow mining algorithm. We assume
that events are logged in temporal order, the event logs
do not contain noise, and the event logs are theoretically
complete.
4.1 Workflow Mining in the Behavior Model
The behavior model shows the behavior perspective of
the workflow. This perspective is the basic and important
of workflow. It is the performance of workflow system.
In the workflow event log table, it contains 3 cases.
We use the algorithm in the 3
rd
section to mine the
behavior model of GBMS.
(1) Definitude the basic relations between the activities
using the D/F-table.
In the log , there are 3 cases, σ
1
={A,B,C,D,E}, σ
2
=
{A,B,D,C,E},σ
3
={A,F,G}.
Then we get the relations between activities:
a>
w
b: A>
w
B,A>
w
F,B>
w
C,B>
w
D,C>
w
D, C>
w
E,D>
w
E,
F>
w
G;
a
w
b: A
w
B,B
w
C,B
w
D,A
w
F,F
w
G,C
w
E,
D
w
E;
a#
w
b: B#
w
F
a//
w
b:C//
w
D, D//
w
C;
(2) Using the α-algorithm to mine the behavior model.
1) T
w
={A, B, C, D, E, F, G};
2) T
i
={A };
3) T
o
={E,G};
4) Xw={(A,B),(B,C),(B,D),(C,E),(D,E),(A,F),(F,
G)};
5) Y
w
=X
w
;
6) P
w
={p(A,B), p(B,C), p(B,D), p(B,D), p(C,E),
p(D,E), p(A,F), p(F,G)} { ∪i
w
,o
w
}；
7) F
w
={(A, p(A,B)), (p(A,B), B), (B, p(B,C)),
(p(B,C), C), (B, p(B,D)), (p(B,D), D), (C,
p(C,E)),(p(C,E), E), (D, p(D,E)), (p(D,E), E),
(A, p(A,F)), (p(A,F), F), (F, p(F,G)), (p(F,G),
G)} { ∪i
w
, o
w
};
8) α (W)={ T w , P w, F w }.
(3) In GBMS, the behavior model is described in the
Figure 2.
Table 1. Workflow event log
Case Activity Status Need Produced Resource
1 A START DOC1 ROLE1
1 A END DOC2
2 A START DOC1 ROLE1
2 A END DOC3
3 A START DOC1 ROLE1
3 A END DOC2
1 B START DOC2 ROLE2
1 B END DOC4, DOC5
3 B START DOC2 ROLE2
3 B END DOC4, DOC5
1 C START DOC4 ROLE4
3 D START DOC5 ROLE4
1 D START DOC5 ROLE5
3 C START DOC4 ROLE5
1 C END DOC6
3 D END DOC7
1 D END DOC7
3 C END DOC6
1 E START
DOC6,
DOC7
ROLE6
1 E END DOC9
3 E START
DOC6,
DOC7
ROLE6
3 E END DOC9
2 F START DOC3 ROLE3
2 F END DOC8
2 G START DOC8 MACHINE1
2 G END DOC9

86 Workflow Mining of More Perspectives of Workflow
Figure 2. The GBMS behavior model
4.2 Workflow Mining in the Information Model
In workflow mining area, there have been no more
researches in informational perspective. In this section,
we introduce the workflow mining in the information
model. The information model is used to describe the
products consumed and produced in the business process
and the relations between the products. In GBMS, the
activity can be executed when it has products and roles
that the states of those must be able to support. So the
information model’s mining is very important and can
show the data flow of the process.
In the workflow event log Table 1, it contains 7
activities and 9 kinds of products. In activity’s START
status, it needs product to run; In END status, it can
produce the new products. So in Table 1 of event log, we
can get the formulas:
9) Prod_Source(A)={DOC1};Prod_Target(A)={ DOC
2, DOC3}
10) Prod_Source(B)={DOC2};Prod_Target(B)={
DOC4, DOC5}
11) Prod_Source(C)={DOC4};Prod_Target(C)={
DOC6}
12) Prod_Source(D)={DOC5};Prod_Target(D)={
DOC7}
13) Prod_Source(E)={DOC6,DOC7};Prod_Target
(E)={ DOC9}
14) Prod_Source(F)={DOC3};Prod_Target(F)={
DOC8}
15) Prod_Source(G)={DOC8};Prod_Target(G)={
DOC9}
Input={Input(A,DOC1),Input(B,DOC2),Input(C,DOC
4), Input(D,DOC5), Input(E,DOC6), Input(E,DOC7),
Input(F,DOC3), Input(G,DOC8)}
Output={Output (A,DOC2), Output (A,DOC3),
Output (B,DOC4), Output (B,DOC5), Output (C,DOC6),
Output (D,DOC7), Output (E,DOC9), Output (F,DOC8),
Output (G,DOC9)}
Figure 3 illustrates the integration of the GBMS
behavior model and data model.
4.3 Workflow Mining in the Organization Model
In GBMS, the organization model is used to define the
government organization; it concludes all kinds of
resources: person, machine, place, etc. In mining this
aspect, a workflow analyst can discover if the participants of
a business process are being used efficiently and
effectively.
Figure 3. The rediscovered of GBMS behavior model
and data model
In the workflow event log Table 1, it contains 7
activities, 6 roles and 1 machine. Activity needs the
resource to support to run in the workflow system. From
Table 1, we can get the formal representation of the
rediscovered organization model:
1) Role (A)={ROLE1}
2) Role (B)={ROLE2}
3) Role (C)={ROLE4}
4) Role (D)={ROLE5}
5) Role (E)={ROLE6}
6) Role (F)={ROLE3}
7) Resource (G)={MACHINE1}
Support ={Support(A,ROLE1), Support(B,ROLE2),
Support (C,ROLE4), Support(D,ROLE5),
Support (E,ROLE6), Support (F,ROLE3), Support
(G,MACHINE1)}
The Figure 4 shows the integration of the GBMS
behavior model, data model and organization model.
This sample is a business process of Beijing Xuanwu
government shown in Figure 5. It is a process of business
application. First the applicant fills in the table of
application, and sends the table for the approval. If the
application is the type of common service, then it is
assigned to Windows’ transaction, and distribute the app
to different leaders to deal with, last it arrives the director
to disposal. If the application is the type of administration
permission, it is directly sent to the administrator to deal
with and recorded into the database of Government. It
was built by GBMS.
Figure 4. The rediscovered of GBMS behavior model,
data model and organization model
A
B
C
D
E
F
G
A
B
C
E
D
F
G
DOC1
DOC
2
DOC
4
DOC
6
DOC
5
DOC
7
DOC
8
DOC
3
DOC
9

Workflow Mining of More Perspectives of Workflow 87
Figure 5. the application business process of GBMS
5. Conclusions and Future Work
Currently, most workflow mining algorithms have the
goal of rediscovering a control flow model. We feel that
this approach limits the scope and utility of workflow
mining; it neglects the important point that workflow is
much more than control flow. So this paper has presented
the workflow mining in more perspective of workflow
and given a full sample to show how to mine the entire
GBMS model from the event log.
In future, we will study the incomplete log with noise
and improve the mining algorithm. This is one research
of the whole project. We also need to validate the mining
model with the model design by the GBMS.
6. Acknowledgement
This project is funded by the Ministry of Science and
Technology, P.R.C.
We should like to thank Beijing Cyber Technology for
supporting the project.
REFERENCES
[1] Workflow Management Coalition, “Workflow management
coalition terminology and glossary,” Technical Report,
WfMCTC–1011, Brussels: Workflow Management
Coalition, 1996.
[2] L. Wang and B. S. Zhou, “The study of enterprise model,”
Computer Engineering and Application, 1002–
8331–(2001)12–0005–05.
[3] B. S. Zhou, H. X, and L. Zhang, “The Principle of Process
Engineering and an Introduction to Process Engineering
Environments,” Journal of Software (supplement), pp.
519–534, August 1997.
[4] B. S. Zhou and S. Y. Zhang, “Visual Process Modeling
Language VPML,” Journal of Software (supplement), pp.
535–545, August 1997.
[5] P. Liu and B. S. Zhou, “The research on the modeling
transformation from GBMS to jBPM,” 2008 International
Conference on Computer Science and Software Engineering.
[6] A. H. Ren, “Research on the Concurrent Software
Developing Method Based on Object Oriented Petri Nets,”
BHU, The school of computer science, pp. 116–128, 2001.
[7] W. M. P. van der Aalst, T. Weijters, and L. Maruster,
“Workflow mining: Discovering process models from
event logs,” IEEE Transactions on Knowledge and Data
Engineering, 16(9), pp. 1128–1142, 2004.
[8] A. J. M. M. Weijters and W. M. P. van der Aalst,
“Process mining: Discovering workflow models from
event-based data,” in: B. Kroose, M. de Rijke, G.
Schreiber, M. van Someren (Eds.), Proceedings of the
13th Belgium-Netherlands Conference on Artificial
Intelligence (BNAIC 2001), pp. 283–290, 2001.
[9] A. J. M. M. Weijters and W. M. P. van der Aalst,
“Workflow mining: Discovering workflow models from
event-based data,” in: C. Dousson, F. Hooppner, R.
Quiniou (Eds.), Proceedings of the ECAI Workshop on
Knowledge Discovery and Spatial Data, pp. 78–84, 2002.