|
Title:
|
A MACHINE LEARNING APPROACH TO IDENTIFYING PROCESS INSTANCE AND ACTIVITY |
|
Author(s):
|
Hongyu Liu , Yunli Wang , Liqiang Geng , Matthew Keays , Nicholas Maillet |
|
ISBN:
|
978-972-8924-97-3 |
|
Editors:
|
Hans Weghorn and Pedro Isaías |
|
Year:
|
2009 |
|
Edition:
|
V I, 2 |
|
Keywords:
|
Process mining, workflow, activity management, machine learning, clustering, classification, conditional random fields. |
|
Type:
|
Full Paper |
|
First Page:
|
3 |
|
Last Page:
|
10 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Current process mining techniques assume structured event log to be in place. The log typically contains explicit
information about events referring to an instance and an activity. However, it is not always the case. For instance, many
structured work or business processes are managed via Email-based communication inside and between organizations. In
this paper, we address the problem of automatically identifying process instances and activities from unstructured event
log such as Emails using machine learning algorithms. The work is focused on two distinct subproblems: 1) Instance
identification is to automatically partition logs into clusters, each of which is a distinct instance. Our approach is to
examine a short time window of the streaming event log to find the clusters, without examining the whole event log. 2)
Identifying the activities in each instance. Our approach is to model activity identification problem as a sequence
labelling task based on Conditional Random Fields (CRFs). We have conducted the experiments on synthetic logs with
different noise types to study the impact on the effectiveness of our approach. |
|
|
|
|
|
|
