P1: Sqe Trim: 6.125in×9.25in Top: 0.5in Gutter: 0.75in
CUUS2079-05 CUUS2079-Zafarani 978 1 107 01885 3 January 13, 2014 19:23
136 Data Mining EssentialsData- Describe methods that can be used to deal with missing data.
- Given a continuous attribute, how can we convert it to a discrete
attribute? How can we convert discrete attributes to continuous ones? - If you had the chance of choosing either instance selection or feature
selection, which one would you choose? Please justify. - Given two text documents that are vectorized, how can we measure
document similarity? - In the example provided for TF-IDF (Example 5.1), the word “orange”
received zero score. Is this desirable? What does a high TF-IDF value
show?
Supervised Learning- Provide a pseudocode for decision tree induction.
- How many decision trees containingnattributes and a binary class can
be generated? - What does zero entropy mean?
- What is the time complexity for learning a naive Bayes classifer?
What is the time complexity for classifying using the naive Bayes
classifier?
Linear separability: Two sets of two-dimensional instances are
linearly separable if they can be completely separated using one
line. In n-dimensional space, two set of instances are linearly
separable if one can separate them by a hyper-plane. A classical
example of nonlinearity is the XOR function. In this function, the
two instance sets are the black-and-white instances (see Figure 5.9),
which cannot be separated using a single line. This is an example
of a nonlinear binary function. Can a naive Bayes classifier learn
nonlinear binary functions? Provide details.
What about linear separability andK-NN? AreK-NNs capable of
solving such problems?
(1,1)
(0,1)(0,0)
(1,0)
Figure 5.9. Nonlinearity of XOR Function.