Performance Evaluation of Support Vector Machine Kernel Functions on Students’ Educational Data Set, Chapter 7.

Simple item page
dc.contributor.authorAbdulsalam, S.O.
dc.contributor.authorGaniyu, R.A.
dc.contributor.authorOmidiora, E.O.
dc.contributor.authorOlabiyisi, S.O.
dc.contributor.authorArowolo, M.O.
dc.date.accessioned2025-10-29T11:48:41Z
dc.date.available2025-10-29T11:48:41Z
dc.date.issued2024
dc.description.abstractSupport Vector Machine (SVM) classifier is currently one of the most popular and extensively used classification algorithm because it produces better classification accuracy performance. However, kernels’ functions parameter optimization during SVM classification process significantly affects its classification accuracy. This research analyses the performance of four SVM kernel functions - Linear Function (LF), Polynomial Function (PF), Radial Basis Function (RBF) and Sigmoid Function (SF) using Grid search method with the intent to determine the best kernel function. The students’ educational dataset used in this research was obtained from the Department of Computer Science in a university in North Central region of Nigeria for a period of 5 years (2009 – 2014). The dataset was a multi-class data comprising records of 153 graduates with 66 predictor variables (sex, age and 64 courses offered by each student) and their final year grade as the class label. The performance of the four kernel functions was evaluated in classification accuracy, sensitivity, specificity, number of support vectors and computation time. The evaluation results of LF, RBF, PF and SF yielded classification accuracies of 79.31, 86.21, 82.76, and 34.5%, respectively. Also, LF, RBF, PF, and SF recorded sensitivities of 71.12, 75.56, 65.56, and 22.20%, respectively. Moreover, LF, RBF, PF, and SF produced specificities of 94.44, 96.34, 95.36, and 74.06%, respectively. Likewise, the number of support vectors generated for LF, RBF, PF, and SF were 78, 80, 82, and 121, respectively. Furthermore, LF, RBF, and PF utilized the same computation time of 0.03 seconds, while SF executed for 0.06 seconds. The findings from the results revealed that RBF kernel is more efficient than the three other kernel functions based on the performance metrics and dataset used in this research. Keywords— Support Vector Machine, Kernels’ functions, Parameter optimization, Grid search method
dc.identifier.urihttps://kwasuspace.kwasu.edu.ng/handle/123456789/6255
dc.language.isoen
dc.publisherCRC Press, Taylor & Francis Group, London - Social Media and Crowdsourcing: Application and Analytics. (pp. 125-149)
dc.titlePerformance Evaluation of Support Vector Machine Kernel Functions on Students’ Educational Data Set, Chapter 7.
dc.typeBook chapter
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
19-Taylor Francis__Performance Evaluation of SVM Kernel Functions on Students' Educational Data Set.pdf
Size:
1.04 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed to upon submission
Description:
Download
Collections
Scholarly Publication