Suppose you are a data scientist in the credit card department of a bank. Your role is to create an automatic program to determine if the bank should approve a client's application for a credit card, or not. You already have millions of examples of customer records labeled with credit card approval history (assume binary labels). Each customer record has a large number (>100) of attributes. Pick the correct strengths and weaknesses of using the basic version of the K-NN classifier to solve this problem. (Select all that apply.) Weakness: if the class proportions for approving credit applications are very imbalanced (eg. in favor of acceptance) the K-NN prediction is likely to be biased toward this majority class. Strength: the K-NN classifier can be updated very efficiently as new labelled examples are added to the dataset Strength: the K-NN classifier only needs to store a small fraction of the entire training set, since it focuses on finding decision boundaries. Weakness: classifying a new instance can be slow, especially if each data point has lots of attributes, because k-NN has to compute the distance to all known instances to find the nearest neighbors. Strength: the K-NN classifier has relatively few parameters to tune. Weakness: K-NN is only effective if the classes are linearly separable (i.e. a hyperplane can be found that splits the positive from the negative examples)