Given a training dataset as a 2d array where rows are examples and columns are features, values associated with those training examples, and some new examples to be predicted, return the value associated with the closest nearest neighbor using a scipy distance metric.

nearest_neighbor.py

import numpy as np
from scipy.spatial.distance import cdist

def nearest_neighbor(samples, targets, samples_to_classify, metric='euclidean'):
    return targets[np.argmin(cdist(samples_to_classify, samples, metric=metric), axis=1)]