1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Process model discovery algorithms based on condensed drop linked list and modified cellular automata
Authors:	Vahideh Naderifar (P61104)
Supervisor:	Zarina Shukur, Prof. Dr.
Keywords:	Computer simulation
Issue Date:	15-Sep-2017
Description:	Process mining technique, which is classified into three classes, including the discovery of processes, conformance of processes and enhancement of processes, uses specialized data-mining algorithms to extract knowledge from the event logs. The event logs contain information about the events of the process. A process consists of multiple cases that relate each event to precisely one case. Events have some attributes, e.g., activity, time, costs, and resource are typical attribute names. Therefore, event logs can be divided from three different perspectives into process, organization, and case perspectives. Process mining suffers from two main challenges related to event logs and process models. The work focuses on representational biasness challenges and balance between "four quality dimensions of conformance checking. Since the existing approaches solve the problems of organization and case perspectives, therefore, this work focuses on control flow of process perspective. Control flow represents relations between all activities in the event log by a graphical model. The construction of common structures such as sequences, choice, loops, non-free-choice, invisible tasks and duplicate tasks is the main difficulty in control flow. Almost all existing algorithms are unable to distinguish the duplicate tasks. For this reason, solving the problem of loops and duplicated tasks are concentrated in this study. Models with loop and duplicated tasks (same activities may execute in different places in the process) increase the complexity of the process models. Linked List and Modified Cellular Automata (MCA) have good potential to mine the process model with loop, duplicate tasks and represent choice, parallelism and sequence and invisible task. This study proposes a two-phased method. In the first phase, a new concept, called as condensate drops, is used as in Linked List for tackling the loops. In order to tackle the duplicated tasks, MCA is utilized with new neighbors and local rules in the second phase. The main objectives of this study are i) to propose Condensed Drop Link List algorithm that produces a control flow process model from event log without loop and refine the created process model using Modified Cellular Automata algorithm for tackling duplicate, sequence, choice, parallelism and invisible tasks. ii) To improve fitness by fitness improvement algorithm which acts on discovery process model of the proposed algorithms and implements the proposed algorithms using two events logs. iii) To carry out conformance checking between a process model and its event log, by validating the process model using four quality metrics and evaluating the process model. The analysis focuses on the fitness, generalization, simplicity and precision of conformance checking. The results demonstrated that the proposed method detected various patterns in the event logs and discovered an efficient process model in terms of the fitness, precision, simplicity and generalization of the mined process model.,Certification of Master's/Doctoral Thesis" is not available
Pages:	258
Call Number:	QA76.9.D343N333 2017 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.