Enhancements of invasive weed optimization algorithm for solving clustering problems

Fatemeh Boobord (P57846)

Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/513257

Full metadata record

DC Field	Value	Language
dc.contributor.advisor	Zalinda Othman, Assoc. Prof. Dr.
dc.contributor.author	Fatemeh Boobord (P57846)
dc.date.accessioned	2023-10-16T04:35:03Z	-
dc.date.available	2023-10-16T04:35:03Z	-
dc.date.issued	2016-12
dc.identifier.other	ukmvital:96518
dc.identifier.uri	https://ptsldigital.ukm.my/jspui/handle/123456789/513257	-
dc.description	Clustering is a data mining tool used to analyse data objects without knowing about the class label. Typically, clustering methods are divided into hierarchical and partitioning. Centre-based techniques, which we focused on in this research, are types of partitioning clustering that search based on the distance of cluster centres. To reduce the cost of distance computation between any two pairs of objects, optimization techniques can be used. Metaheuristics are optimization methods that are categorized into S-metaheuristic and P-metaheuristics, and both categories are usually used for data clustering. Among the optimization approaches, the Invasive Weed Optimization or IWO is a P-metaheuristic algorithm. The best advantage of IWO is using fit and unfit solutions in the searching process as this gives more chance to find the nearer solution to global optima. The advantage of IWO makes it different from other P-metaheuristics. Previous works have shown the effectiveness of the IWO algorithm in data clustering. However, similar to other optimization algorithms, the IWO has problems of trapping to local optima and premature convergence. The increase of data dimensions limits the capability of existing IWO to handle clustering in varied sizes of data. The aim of this study is to propose two enhancements of the IWO to handle local optima and varied sizes of data problems. Firstly, an enhanced hybrid of the IWO and Particle Swarm Optimization (PSO) is proposed to handle premature convergence and escape from the local minimal. The original RGIP hybrid algorithm is improved by using the PSO for generating an initial population called the PGIP. The performance of PGIP is measured using the internal quality measure (SSE) and external quality measure (accuracy of clusters). The experimental results showed that the PGIP outperforms other methods consistently. In the second enhancement, two level IWO based algorithms are proposed (IWO+IWO) which are IWO based feature selection algorithm and integrated IWO clustering algorithm. The IWO based feature selection employed the Principle Component Analysis (PCA) to specify the number of reduced feature as a parameter to reduce data size. Then, the IWO clustering algorithm is employed again to cluster the data. The proposed IWO+IWO algorithm is tested upon various sizes of dataset and it has the best performance in terms of internal quality measure when compared with other methods. The IWO based feature selection method outperforms the PCA based feature selection, increasing the number of features and the capability of IWO+IWO in clustering varied sizes of data.,Certification of Master's/Doctoral Thesis" is not available
dc.language.iso	eng
dc.publisher	UKM, Bangi
dc.relation	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat
dc.rights	UKM
dc.subject	Data clustering
dc.subject	Cluster analysis
dc.title	Enhancements of invasive weed optimization algorithm for solving clustering problems
dc.type	Theses
dc.format.pages	142
dc.identifier.callno	QA278.55.B645 2016 3 tesis
dc.identifier.barcode	002719(2017)
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:

File	Description	Size	Format
ukmvital_96518+SOURCE1+SOURCE1.0.PDF Restricted Access		188.48 kB	Adobe PDF	View/Open

Show simple item record Recommend this item