QUESTION 1

1. Which of the following statements about k-NN is FALSE?

   	A.	To classify or predict a new record, the k-nearest neighbormethod relies on finding similar records in the training data.
   	B.	The k-nearest neighbor method cannot be used to predict aquantitative variable--it can only be used to classify or predictthe outcome of a categorical variable.
   	C.	Both the k-nearest neighbor method and logistic regression canbe used to predict a categorical variable.
   	D.	The k-nearest neighbor method determines neighbors by measuringdistance between records based on the values of their predictorvariables.

6 points

QUESTION 2

1. Which of the following statements is FALSE about k-NN?

   	A.	The k-nearest neighbor method determines neighbors by measuringdistance between records based on the response variable.
   	B.	In k-NN, Euclidean distance is the most popular method ofcomputing distance between each record to be predicted andevery record in the training set.
   	C.	k-NN uses a majority decision rule to classify a new record,where a record is classified as a member of the majority class ofthe k neighbors.
   	D.	How to choose k in k-NN? Answer: choose the k that has the bestclassification performance measured by % of error.

6 points

QUESTION 3

1. Which statement about k-NN is FALSE?

   	A.	k-NN uses a majority decision rule to classify a new record,where a record is classified as a member of the majority class ofthe k neighbors.
   	B.	How to choose k in k-NN used for classification? Answer: choosethe k that has the best classification performance measured by % oferror.
   	C.	k-NN is a "lazy learner" in the sense that the time-consumingcomputation is deferred to the time of prediction which prohibitsits use for real-time prediction of large number of recordssimultaneously.
   	D.	In k-NN method, Euclidean distance is the least popular methodof computing distance between each record to be predicted andevery record in the training set because it is computationally mostdemanding.

6 points

QUESTION 4

1. Identify the statement about k-NN that is NOT true.

   	A.	Higher values of k in k-NN provide smoothing that reduces therisk of ovrefitting due to noise in the training data.
   	B.	How to choose k in k-NN used for classification? Answer: choosethe k that has the best classification performance measured by % oferror.
   	C.	How to choose k in k-NN? Answer: choose the largest k whichcoincides with the naïve rule.
   	D.	One difficulty of k-NN method is that the computational burdenand time to find the nearest neighbors in a large training set canbe prohibitive.

Answer rating: 100% (QA)

Q1 B Q2 A Q3 C Q4 The knearest neighbor method cannot be used to predict a quantit... View the full answer