In this assignment you will be implementing a straightforward classification algorithm known as the k-nearest neighbor (k-NN) classifier.  Your implementation should accept two data files as input (both of which are posted in Blackboard together with this assignment):

1. A training dataset of class-labeled feature vectors

2. A test dataset of unlabeled feature vectors

For simplicity, you do NOT need to normalize the data for this exercise.

The training data contains examples of three different types of Iris (a type of flower), with each example having a class-label of either versicolor, virginica or setosa. Each example has four (numeric) features:  Sepal Length, Sepal Width, Petal Length and Petal Width. Your classifier must examine each unlabeled example in the test set and classify it as one of the three classes of Iris.  The classification will be based on an unweighted vote of its k nearest examples in the training set.

To determine nearest neighbors, you should measure all distances using regular Euclidean

distance:

If two or more classes receive the same (winning) number of votes, break the tie by choosing the class with the lowest total distance from the test point to its voting examples.   For example, if class X has two votes  (at distances  1 and 3) and class Y has two votes  (at distances 2 and 4), then your algorithm should predict class X since (1+3)<(2+4).