MLII - Nearest Neighbor Methods

What is the main idea behind 'Instance based' classification?

• Select the class of the most similar instance

• Select the class of the least similar instance

• Select the class of the most frequent instance

• Select the class of a random instance

What is the purpose of distance measures in distance-based machine learning methods?

• To find items that are randomly distributed

• To find items that are similar based on distance

• To find items that are different from each other

• To find items that are not mixable

What is the function of Term Frequency / Inverse Document Frequency in a vector space?

• To calculate the magnitude of vectors

• To determine the direction of vectors

• To measure the distance between vectors

• To identify the importance of a word in a document collection

How is Cosine similarity used in Information Retrieval?

• To classify documents

• To rank retrieved documents

• To delete irrelevant documents

• To create new documents

What is the purpose of the k-nearest neighbors algorithm?

• To find neighbors with the same age

• To find similar items in close proximity

• To find the neighbors with the highest salary

• To find the most distant neighbors

What is the advantage of Weighted Voting in kNN?

• It only considers the closest neighbor

• It introduces enough variation to prevent ties

• It ignores the distance between neighbors

• It always counts all neighbors equally

In Collaborative Filtering, how are predictions made for a new user?

• By ignoring the ratings of other users

• By averaging the ratings of the closest neighbors

• By assigning the user to the class with the most ratings

• By randomly assigning ratings

What is a limitation of KNN in terms of interpreting the model?

• It is difficult to interpret beyond the similarity between neighbors

• It provides detailed insights into the decision boundary

• It highlights the importance of each predictor

• It simplifies the target function

When is KNN most suitable to use?

• When classification accuracy is not important

• When there are more than 20 features

• When feature values are not scaled

• When there is limited training data

What is an advantage of KNN over more complex models?

• It is easily fooled by irrelevant attributes

• It requires extensive optimization and training

• It is fast at query time

• It can learn complex target functions