How is linear discriminant analysis (LDA) different from quadratic discriminant analysis (QDA)? (select ALL that apply)
LDA has a linear discriminant function
QDA has more parameters to estimate
LDA assumes the same covariance matrix for the distribution of X across all classes
QDA assumes arbitrary distribution for each class.
Discriminant analysis can be used to answer questions such as:
How much of the variation in sales can be explained by advertising expenditures, prices, and level of distribution?
In terms of demographic characteristics, how do customers who exhibit store loyalty differ from those who do not?
What are the distinguishing characteristics of consumers who respond to direct mail solicitations?
Both b and c are correct.
Which of the followings are correct about the k-NN algorithm?
k-NN algorithm does more computation on train time rather than test time.
k-NN algorithm does more computation on test time rather than train time.
In k-NN algorithm, when you increase the k, the bias will be increased
In k-NN algorithm, when you increase the k, the bias will be decreased.
How we can avoid the overfitting in Decision Tree?
Stopping the tree growth
Pruning the full grown tree
Both of above
None of above
Consider a 2-dimensional dataset with features x1 ∈ [0, 5], x2 ∈ [0, 4] and label y ∈ {0, 1}, and we fit a classification tree. Which of the following are correct? (select ALL that apply)
On the feature space, the classification tree will cut the large rectangle [0, 5] × [0, 4] into small rectangles.
On the feature space, the region corresponding to the leaves of the classification tree are not necessarily rectangles.
In order to make the classification tree method work well, we have to assume distributional assumption on x1 and x2.
We can use the classification tree to predict the class probabilities of y.
Which of the following are model free algorithms? (select ALL that apply)
LDA
Decision Trees
QDA
k-NN
Which of the following are true for decision trees? (select ALL that apply)
This algorithm can only be applied when the features x are continuous
This algorithm can produce nonlinear classification boundaries
A larger size of decision trees means larger bias and smaller variance
This algorithm can handle missing values without missing data imputation
