Firefly algorithms with probabilistic neural network for classifications problems

Mohammed Hasan Ahmad Alweshah (P56200)

Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/513348

Full metadata record

DC Field	Value	Language
dc.contributor.advisor	Salwani Abdullah, Assoc. Prof. Dr.
dc.contributor.author	Mohammed Hasan Ahmad Alweshah (P56200)
dc.date.accessioned	2023-10-16T04:35:45Z	-
dc.date.available	2023-10-16T04:35:45Z	-
dc.date.issued	2013-11-11
dc.identifier.other	ukmvital:119535
dc.identifier.uri	https://ptsldigital.ukm.my/jspui/handle/123456789/513348	-
dc.description	Classification is a data mining task that assigns items in a collection to predefined categories or classes which is also referred as a supervised learning. The goal of classification is to accurately predict the target class for each case in the data. Classification becomes a significant optimization research area such as in financial, organizational, and medical problems. Literature review shows that many algorithms including statistical and machine learning have been successfully used to handle classification problems in different areas, but their performance varies considerably. One of the most well-known algorithms is the Neural Network (NN). Even though the NN is effective in addressing a wide range of problems, to date no specific NN approach has been found that can ensure optimal solution in solving classification problems. Although there are some approaches are more effective than others, but no single classifier that offers the best performance for all tested datasets. This motivated the investigation of firefly algorithms (FA) that to date, it has not been applied to classification problems. FA is based on flashing characteristics of fireflies in communicating and finding mates. Some of the important challenges include: finding the most appropriate weight parameter of the classifier through the implementation of the population-based approaches, attaining a balance between the exploration and exploitation processes by employing hybridization methods, obtaining fast convergence by controlling the random movement and by generating good initial solutions. In order to tackle the listed challenges above, the research presented in this thesis provides effective approaches in optimizing the weight of the classifier thus lead to higher accuracy for classification problems. This has been achieved mainly via a modification of population-based approach i.e., the FA. The research first highlights the use of FA to optimize the weight parameters of the Probabilistic Neural Network (PNN) for classification problems in order to improve the classification accuracy. Secondly, in order to obtain a balance between the exploration and exploitation during the search process, the FA is hybridized with a Simulated Annealing (SA) local search algorithm (coded as SFA). Thirdly, a Lévy flight (LFA) is hybridized with an aim to speed up the convergence speed by controlling the random movement within the SFA (coded as LSFA). To further improve the performance of the LSFA, a constructive heuristic that generate good initial solutions for the LSFA has been investigated with an intention to see the effect on the convergence speed and the final objective values (with respect to the classification accuracy) (coded as ILSFA). The proposed approaches are tested on 11 standard benchmark data sets. Initial experiments show that FA is able to tune the weight parameter of the PNN that leads to better classification accuracy in comparison to PNN in isolation. The obtained results are further improved significantly by SFA, followed by LSFA and ILSFA. Overall comparisons indicate that ILSFA works well across all datasets. This is a new approach in the classification arena and it represents an approach that outperforms the current state-of-the-art on most of the tested benchmark datasets.,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	Firefly algorithms
dc.subject	Network
dc.subject	Neural networks (Computer science)
dc.subject	Universiti Kebangsaan Malaysia -- Dissertations
dc.subject	Dissertations, Academic -- Malaysia
dc.title	Firefly algorithms with probabilistic neural network for classifications problems
dc.type	Theses
dc.format.pages	187
dc.identifier.callno	QA76.87.A459 2013 3 tesis
dc.identifier.barcode	002748 (2013)
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:

File	Description	Size	Format
ukmvital_119535+SOURCE1+SOURCE1.0.PDF Restricted Access		1.13 MB	Adobe PDF	View/Open

Show simple item record Recommend this item