Descision Boundary tool for SVM classifier

• Classifiers seperate classes by mapping classes to a probabilty space
• In understanding how well a classifier is working it helps to visualize the descision boundary created
• Changing the parameters of the classifier changes the fit. Understanding the parameters is key to using the classifier with sucess.
• The Support Vector Machine (SVM) is known as a large margin classifier and uses two parameters C and gamma to control the fit

Below is a plot of the example dataset used by this tool Visually we can see that the two classes are not seperable linearly and it also has some outliers.

• When using the descision boundary tool and adjusting the C parameter we can see that the SVM classifier attempts to classify all points (including outliers) correctly to reduce the misclassification error when using high C values. Where as a low C value produces a simple decision boundary with higher misclassification error. The C acts as a regularization parameter and can be used to prevent overfitting the data
• The parameter gamma determines the radius of influence of a training point and large values lead to overfit and small values result in simple decision boundary.

• Learning methods are effective tools in supervised learning but the user needs to know how the parameters affect the outcome
• Educative lools like this one provide an intuitive learning as to how the parameters influence the model