ML-Classification

This report shows various well-known methods of classification.

Including linear classifier from scratch, linear classifier with least-squared manner, voted perception and SVM.

Finally, I make a comparison by calculating accuracy between above mentioned method.

Content

1. All codes used are in "code" file

   • Linear Classifier from scratch : LC

   • Linear classifier with least-squared manner : LS

   • Voted perceptron

   • SVM (Hard margin) : HardSVM

   • SVM (Soft margin) : SoftSVM

   • SVM by sklearn : sklearn

2. Figures of performance with different C are in "figure" file

   • data - Performance with different C_data

   • crx - Performance with different C_crx

3. Given datasets "data.csv" and "crx.csv" are im "Given dataset" file

Result

ACCURACY OF EACH METHOD

	LC	LS	VP	SVM(Hard)	SVM(Soft)	SVM(sklearn)
data	0.9156	0.9649	0.9104	1.0	0.9824	0.9591
crx	0.6937	0.5345	0.6217	0.6018	0.8760	0.8827

COMPARISON OF MARGIN

	LC (Scratch)	SVM (Hard)
data	0.0001	4.1371e-05
crx	0.0017	0.4905

Effective weighting value C