Extracting Connection Types in Process Models Discovered by Using From-to Chart Based Approach - Developing Concepts in Applied Intelligence - page 59

Extracting Connection Types in Process Models
Discovered by Using From-to Chart Based Approach
Eren Esgin 1 and Pinar Senkul 2
1 Middle East Technical University, Informatics Institute,
065531 Ankara, Turkey
2 Middle East Technical University, Computer Engineering Department,
065531 Ankara, Turkey
[e-mail] [e-mail]
Abstract. Although contemporary information systems are intensively utilized
in enterprises, their actual impact in automating complex business process is
still constrained by the difficulties coincided in design phase. In this study, a
hybrid data analysis methodology to business process modeling that is based on
using from-to chart is further enhanced to discover connection types and
include in the process model. From-to chart is basically used as the front-end to
figure out the observed transitions among the activities in event logs. The
derived raw patterns are converted into activity sequence on from-to chart by
using Genetic Algorithms. Finally a process model including AND/OR
connection types is constructed on the basis of this activity sequence.
Keywords: From-to Chart, Business Process Modeling (BPM), Process
Mining, Connection Types, Genetic Algorithms (GA), Event Logs.
1 Introduction
In conventional process modeling, a reference model for the target process is
constructed. However, sometimes the actual enactment may deviate from the proposed
reference model due to the reasons such as subjectivity of the domain experts or
incomplete design phase. Process mining is proposed a remedy to these discrepancies
by mining interesting patterns from the event logs and discovering the business process
model automatically [5,6,7,8,9,10] . Unlike contemporary process modeling approaches,
process mining is not biased or restrictive by normative perceptions [14] .
A new approach for process mining based on using from-to chart is given in [2]. In
this approach, from-to chart, which is fundamentally used in tracing material handling
routes on the production floor [4] , is adapted to the business process modeling (BPM)
domain and is used for analyzing the event logs. The transitions among activities
occurred in process instances are kept and analyzed in a from-to chart in order to
check if there exists any specific order of the occurrences for representing in process
model [2] . This approach is further improved in [3] in such a way that the runtime
complexity of the approach is improved by adopting Genetic Algorithms (GA) [11] .
In this work, we extend the work in [3] by extraction of AND/OR connection types.
Through this extension, the type of connections are discovered for each parallel