Results

Linear Kernel:

The confusion Matrix for the linear Kernel for Cost = 10.

The accuracy achieved is just 46.64% for Cost=10. Suggesting that the linear kernel might not be the right fit for this data.

By decreasing the Cost to 5, results in a lower accuracy of ~38%.

Polynomial Kernel:

The Confusion Matrix for the polynomial Kernel for Cost = 5.

For Polynomial kernel, with cost equals to 5, we get a better accuracy than the linear kernel with accuracy of 56.33%

With cost equals to 15, we get a slightly better accuracy than the than before of 57.29%

Radial Basis function(rbf) Kernel:

The Confusion Matrix for the RBF Kernel for Cost = 3.5

The accuracy for RBF kernel is almost comparable with polynomial kernel.

The accuracy for RBF kernel is 57% for cost equal to 10.

Out of all three kernels used, we haven't got good accuracy, suggesting that it might require more hyperparameter tuning to achieve better accuracy and also might require a different kernel other than the three used above. From the three, RBF and Polynomial kernels gave almost comparable accuracy around 57%. Thus both were better than Linear Kernel.

Page updated

Report abuse