support vector machine Flashcards
what is SVM
classification technique used for predicting binary outcome.
what does SVM require d
labelled and balanced dataset.
SVM is a _____ technique
black box technique
we will not be able to know which variables has a greater influence on the outcome nor how much is the influence of each variable on the outcome.
What factors affect the outcome and how much is the effect of each factor?
What factors affect the outcome and how much is the effect of each factor?
how does SVM work
*Support vector machine (SVM) learns a hyperplane to classify data into 2 classes
*A line to separate blue & red dot.
*The balls represent data set.
*The red and blue colors represent 2 classes.
*The stick represents the simplest hyperplane which is a line.
*SVM figures out the kernel function for the hyperplane.
*If a new red ball comes, we know where the new ball should be placed
gets more complex when data points are not linearly separable
what is the kernel trick
1.Transfer (lower dimension) Input Space into (higher dimension)
Feature Space: 2D -> 3D
2.A Kernel function is used for this transformation
3.Find a plane in this Feature Space to separate balls
what is a kernel function
A kernel function maps pairs of datapoints onto their inner products.
A feature space based on a kernel function has one dimension for every pair of datapoints.
Mathematical minimisation can then be used to find the max-margin hyperplane in the feature-space.
The effect is to identify a non-linear (curved) boundary in the original data space.
where does SVM work in ?
SVM works only in a real-valued space. For a categorical attribute, we need to convert its categorical values to numeric values.
advantage of kernel based framework
Kernel-based framework is powerful and flexible
true or false
SVM works very well in practice, even with very small training sizes
true
SVM does only does ____ class classification
2
Multi-class SVM must combine _____
2 class SVMs
drawback of svm
*Learning from training datasets may take a very long time for large-scale problems
accuracy formula
(TP + TN) / (P + N)
weakness of accuracy formula
FP and FN are absent from the accuracy formula, this is where its weakness lies
