Model Evaluation

a) Let us consider a modified version of K- Nearest neighbor which simply functions like a lookup table. That is, given a test example to classify, if the test example is present in the training dataset the classifier outputs its label as the one in the training dataset. If the test example is absent in the training dataset, the classifier outputs its label as -ve. Given a training dataset with no duplicate entries and 50% of the dataset is of the +ve class and 50% is of the ve class, what is the average leave-one-out cross validation error.

b) Given the following training and testing datasets, what is the precision and recall of our classifier in the test data w.r.t to the +ve class.

Feature-1 X1 X2 X3 X4 X5 X6 Feature-2 Y1 Y2 Y3 Y4 Y5 Y6 Class Training data Feature-1 X1 X2 X3 X10 X11 Feature-2 Y1 Y2 Y3 Y10 Y11 Class Testing data Feature-1 X1 X2 X3 X4 X5 X6 Feature-2 Y1 Y2 Y3 Y4 Y5 Y6 Class Training data Feature-1 X1 X2 X3 X10 X11 Feature-2 Y1 Y2 Y3 Y10 Y11 Class Testing data