A comparative study of support vector machine and logistic regression article · January 021 citations reads 11 authors
Yüklə 395,89 Kb. Pdf görüntüsü
|
- Bu səhifədəki naviqasiya:
- 4.0 Conclusion
|
Page 90
(SVM_LK), while the LR recorded the least media. The findings showed that the SVM using the Gaussian Kernel was a better classifier than SVM using the linear Kernel and the LR. In fact, the result of ANOVA test in R (Appendix C.3) shows that at a p-value of 0.0152 and 0.0119 respectively, the null hypothesis that the performances of the three classifiers do not essentially differ from each other and was rejected. This concludes that the SVM using the Gaussian kernel is preferred over other classifiers considered in the present study for classification problem, given the datasets involved in the analysis. Table 4.1 : Accuracy rates of SVM and LogReg on separate datasets. 4.0 Conclusion In conclusion, and particularly based on Table 4.1 and Figure 4.1, showed that the SVM using the Gaussian Kernel was a better classifier than the other classifiers considered in the study. This S/nos Names of Dataset Dimensions SVM Linear Kernel SVM Gaussian Kernel LOGREG 1 Bupa 345 0.6442 0.79615 0.3653 2 Saheart 462 0.7266 0.7618 0.7553 3 Haberman 306 0.6848 0.7065 0.5716 4 Appendicitis 106 0.8438 0.8125 0.8438 5 Banana 5300 0.5547 0.8969 0.5107 6 Heart 270 0.8272 0.8821 0.8271 7 Titanic 2201 0.7932 0.7518 0.5541 8 Heart Disease 303 0.7362 0.7961 0.6780 9 Monk 432 0.7308 1.0000 0.7077 10 Cmc 1473 0.4886 0.5068 0.4909 11 Hayes_Roth 160 0.5750 0.7164 0.5750 12 Tae 151 0.8666 0.84444 0.6304 13 Lung Cancer 56 0.6000 0.7000 0.6000 14 Abalone 4177 0.96807 0.9689 0.9633 15 Pima 768 0.6782 0.6913 0.550 16 Immunotherapy 90 0.8519 0.7407 0.8148 17 Air Quality 9358 0.7436 0.9342 0.6750 18 Infant Mortality 164 5 0.5102 0.5714 0.4286 O RIENTAL JOURNAL OF SCIENCE & ENGINEERING VOL -2, ISS-1, FEB - 2021 www.ojse.org ojse©2019 Page 91 position is corroborated by the findings of the one-way Analysis of Variance and the Post Hoc test using the Bonferroni test. The test clearly rejected the null hypothesis that the performances of the three classifiers do not differ from one another. This further concludes that with respect to the datasets used in the study, that the SVM using the Gaussian Kernel is a preferred classifier than the LR. Yüklə 395,89 Kb. Dostları ilə paylaş:
|