000			04088ntm a2200337 i 4500
999			_c100808 _d100814
003			MY-KuUP
005			20251125110857.0
006			t||||fr|||| 000 0
007			ta
008			240524t20242024my a|||fr|||| 000 0 eng d
020			_aTHE0009845 (Local) _qHardback
040			_aUMPSA _beng _cUMP _erda
090			_aFTKEE .R38 2024 r Thesis
100	1		_aRavindran A/l Nadarajan, _eauthor.
245	1	0	_aOptimisation of support vector machine hyperparameters using enhanced artificial bee colony variant to diagnose breast cancer / _cRavindran A/l Nadarajan
264		1	_aKuantan, Pahang : _bUMPSA, _c2024
264		4	_c©2024
300			_axiii, 181 pages : _billustrations (some color) ; _c30 cm. + _e1 CD-ROM
336			_2rdacontent _atext
337			_2rdamedia _aunmediated
338			_2rdacarrier _avolume
347			_2rda _atext file _bPDF
500			_aFaculty of Electrical & Electronics Engineering Technology
502			_aThesis (Master of Science) -- Universiti Malaysia Pahang – 2024
504			_aIncludes bibliographical references
520	3		_aBreast cancer is the leading cause of cancer-related deaths among women globally. Early detection through methods such as mammography and biopsy are crucial in reducing cancer-related mortality. However, these methods are vulnerable to inaccuracies due to human intervention in diagnosis. Machine learning offers a solution to these challenges, with support vector machines (SVM) being a popular choice for breast cancer diagnosis given its strength in binary classification, which suited well with the dataset used in this thesis. The performance of SVM can be affected by hyperparameters, which are kernel scale and known as gamma and regularization parameters (C). A metaheuristic algorithm is introduced to optimise the hyperparameters. This algorithm named JAABC5ROC is the enhancement of Artificial Bee Colony (ABC) variant, JA-ABC5 by combining with Rate of Change (ROC)\. The combination of these two aspects can assist to balance and enhance the exploration and exploitation capability. The combination of these two aspects can assist to balance and enhance the exploration and exploitation capability. Before using the JAABC5ROC as an optimizer for the SVM, a total of 10 benchmark function were used to determine its performance assessment 5 common benchmarks which are (Shows Rosenbrok, Sphere, Step and RS Schwefel Ridges and RS Zekhelip) and 5 CEC2017 benchmarks which are (Shifted and Rotated Zakharov Function, Hybrid Function 01, Composite Function 08, Composite Function 09 and Composite Function 10). The JAABC5ROC is chosen for its better optimisation from the benchmark result. The hyperparameter of the SVM is optmised with 100,500,1500 and 3000 iteration and record the value. The criterion for its performance is based on the specificity, sensitivity and accuracy. The SVM-JAABC5ROC (SVM optimised by JAABC5ROC) was compared with three optimisers (Grid search, Random Search and Bayesian Optimisation) and compared with 5 classifiers (Artificial Neural Network (ANN), K-Nearest Neighbour (KNN), Naïve Bayes, Logistic Regression and Decision Tree). Both optimisers and classifiers employed the Wisconsin and Mammography dataset. The Wisconsin dataset has 31 features and 569 patients for Mammography dataset with five features and 962 patients. The results revealed that the SVM-JAABC5ROC has an accuracy of 98.6% for Wisconsin Dataset and 85.63% in Mammography Dataset when compared with the optmiser. For the classifier comparison the accuracy of SVM-JAABC5ROC is 94.9% for Wisconsin Dataset and 85.63% for mammography dataset. The SVM-JAABC5ROC outperformed all three optimizers and five classifiers in terms of accuracy for both the Wisconsin and Mammography datasets. This indicates that the SVM-JAABC5ROC is a highly effective model for classification tasks on these datasets.
610	2	0	_aFaculty of Electrical & Electronics Engineering Technology _xDissertations
650		0	_aUniversities and colleges _xDissertations
650		0	_aTheses _xDissertations
942			_2lcc _cTHESIS