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

Title:	Dynamic Ant Colony System With Three Level Updates (DACS3) Based On Individual Ant Behaviour
Authors:	Helmi Md Rais (P36134)
Supervisor:	Zulaiha Ali Othman, Associate Professor Dr.
Keywords:	Dynamic Ant Colony System Dynamic Ant Colony System With Three Level Updates Dynamic Ant Colony System With Three Level Updates DACS3 DACS3 Three Level Updates On Dynamic Ant Colony System Ants--Behavior--Mathematical models
Issue Date:	20-Apr-2013
Description:	Currently, Ant Colony Optimization (ACO) metaheuristic is one of the prominent techniques applied in TSP and VRP solution. It is based on the cooperation of a complex society of ants through a chemical substance called pheromone. Various versions of ACO metaheuristic have been developed which differ in two conditions: decision making on next movement and updating pheromone at two level updates. However, the ACOs' still not yet reached the standard benchmark solution. Moreover, previous researchers less focuses on exploiting experiences of individual ants in updating pheromone level. Embedding such behaviour is believed can improve the ACO algorithm's performance. Therefore, the purpose of this research is to propose a Dynamic Ant Colony System with Three Level Updates (DACS3) algorithm that embedding Malaysian House Red Ants behaviour into current ACS. The algorithm consists of three levels of pheromone updating rules such as local, intermediate and global pheromone updates. The individual behaviour of ants plays an active role at intermediate level, which offers extra exploration to global worse and global best. Research methodology consists of five phases ie: literature reviews, observation of the Malaysian House Red Ants, building the fundamental principle of DACS3 algorithm based on the result of the observation and testing the effectiveness of the proposed algorithm with the reconstructed algorithms ie: ACS and DACS. The proposed algorithm was further investigated and improved using experimental methodology developed in three stages such as basic, enhance and hybrid. The algorithm performance was tested using datasets from two domains ie: TSP and Capacitated VRP. The performance of algorithm is measure based on the quality of solution (the shortest distance), time taken (second) to reach the solution and searching performance. The algorithm searching performance was tested using five methods of quantitative measurements such as Log Graph, Non-parametric (Mann-Whitney) significant test, Paired-Samples significant T-test, ROC Curve significant test and Descriptive test. Generally, DACS3 enables in finding better solution in term of the quality of solution (0% - 5%) and faster time (4% - 90%) compared to ACS and DACS. The result also shows that DACS3 has an increased sensitivity in making decision for the next movement to produce a better solution but has a reduced capability in searching performance at the beginning for large data. As the conclusion, by embedding a simple ant behaviour has improved ACO's algorithm searching performance and its sensitivity or effectiveness. However, the embedding has reduced its searching performance capability at the beginning of the search for large data due to increased selective activities resulting from an increase of available information.,PhD
Pages:	316
Call Number:	QA402.5 .H443 2013 3
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.