LIT Presentation Amna
-
Upload
amna-shifia-nisafani -
Category
Documents
-
view
224 -
download
0
Transcript of LIT Presentation Amna
-
8/6/2019 LIT Presentation Amna
1/14
www.rclit.com
Activity Sequence Prediction
for Better Scheduling in BPM
Activity Sequence Prediction
for Better Scheduling in BPM
Project 0-0
Presenter : Amna Shifia Nisafani
Date : 2011. 7. 15
-
8/6/2019 LIT Presentation Amna
2/14
In this project, we demonstrate how to predict the activity sequencein BPM.
In classical scheduling :
all resources and all activities are given.
There is no uncertainty in the behavior of resources and activities.
However, in BPM scheduling : Information about resources and activities are not always provided
There is uncertainty in the behavior of the resources and activities.
Therefore, it is necessary to predict the incoming activities that
highly to be executed in order to improve BPM scheduling
performance.
-
8/6/2019 LIT Presentation Amna
3/14
This study proposes a modeling approach to predict highly incurred
activities for better scheduling in BPM.
-
8/6/2019 LIT Presentation Amna
4/14
A sequence is the order or path of activities to be performed in a process.
Activity Execution Sequence (AES)
an ordered list of activities that are likely to be executed, and changes
dynamically while a process is being carried out.
-
8/6/2019 LIT Presentation Amna
5/14
Markov Chain is a model that generate sequences in which the probability of symbol depends only on the previous symbol
In a Markov Chain model, the model is defined by
a set ofstates, Q, which emit symbols, and
a set oftransitions between states
Each transition has an associated probability,pkl, which represents the conditional probability of going to activity lin the next activity, given that the
current activity is k.
-
8/6/2019 LIT Presentation Amna
6/14
-
8/6/2019 LIT Presentation Amna
7/14
The transition probabilities can also be written as a transition matri
x, P={pkl}. The value ofpklcan be obtained aspkl= Pr(al|ak).
-
8/6/2019 LIT Presentation Amna
8/14
-
8/6/2019 LIT Presentation Amna
9/14
Example of sequence prediction
-
8/6/2019 LIT Presentation Amna
10/14
We apply a simple BP. We use Java to develop simulation environment.
We have considered some simple dispatching rules such as First In First Out(FIFO), Shortest Processing Time (SPT), Earliest Due Date (EDD) andOperation Due Date (ODD) in order to capture the scheduling process in a
system. Mean flow time : FIFO vs. SPT Mean lateness : EDD vs. ODD Without sequence prediction: FIFO and EDD With sequence : SPT and ODD
-
8/6/2019 LIT Presentation Amna
11/14
0
10
20
30
40
50
60
70
Mean Flow Time
w/o uncertainty reduction
w/ uncertainty reduction
Replication #
(Thousand)w/o sequence prediction
w sequence prediction
330
350
370
390
410
430
450
1 2 3 4 5 6 7 8 9 10
Mean Lateness
w/o uncertainty reduction
w/ uncertainty reduction
Replication #
w/o sequence prediction
w sequence prediction
Comparison of Scheduling with and without sequence prediction
-
8/6/2019 LIT Presentation Amna
12/14
In this study, we propose a method to forecast the occurrence of activit
ies in BPM.
We demonstrate out method in two scheduling environment : with an
d without activity sequence prediction.
It is evident that the better forecast lead the better scheduling performance.
-
8/6/2019 LIT Presentation Amna
13/14
Some research gaps remain to be filled, for instance, the means by which the accuracy of our sequence prediction algorithm
, which has a direct influence on the process completion time
, can be increased.
Other pending research includes the study of approaches for
dealing with large data logs that can diminish the performance of the mining processes.
-
8/6/2019 LIT Presentation Amna
14/14
THANK YOU