$($ Q $1$ A $)$

Short Answers

$($a$)$ What are the three types of classification errors? Explain the solutions to each of these three errors.

$($b$)$ For Nearest Neighbor classification, missing data in training set is a problem. Explain how to solve this problem.

$($c$)$ Na$$ve Bayes classification assumes the discrete training dataset, how to deal with continuous data? Explain your method in details.

$($d$)$ How to use SVM to classify training data if decision boundary is not linear? Explain using an example in details.

$($Q$1$B$)$

True or False $(10$ points$)$:

$($a$)$ Even if you have training data with same support vector points, the decision boundary might change due to different learning processes.

$($b$)$ In SVM$,$ the decision boundary is determined only by support vectors.

$($c$)$ The association analysis will generate the same frequent itemsets and strong association rules no matter that a specific item occurs once or multiple times in an individual transaction.

$($d$)$ In association rule mining the generation of the frequent itermsets is the computational intensive step.

$($e$)$ The perceptron learning algorithm can generate linear hyperplanes for nonlinearly separable problems.