.. battelle.knn

.. currentmodule:: battelle.knn

K-nearest Neighbors Classifier subpackage (battelle.knn)
=========================================================

.. autosummary::
   :toctree: generated/

   KNearestNeighborsClassifier

   .. automethod:: 

      KNearestNeighborsClassifier.predict
      KNearestNeighborsClassifier.train

This module allows the user to create and train k-nearest neighbor classifiers.

Here is a simple example of binary classification:

.. code-block:: python

   import numpy as np
   import matplotlib.pyplot as plt

   from battelle.knn import KNearestNeighborsClassifier

   neigh = KNearestNeighborsClassifier()

   x = np.random.random(size=(300,2)) - 0.5
   x = [list(xi) for xi in x]

   y = np.array([np.sign(np.prod(k)) for k in x])

   neigh.train(x,y)

   # Plot training data
   plt.subplot(1,2,1)
   for i, xi in enumerate(x):
       if y[i] == 1:
           plt.plot(xi[0], xi[1], 'xr')
       else:
           plt.plot(xi[0], xi[1], 'xb')
   plt.title("Testing data")

   # Plot neural net result
   plt.subplot(1,2,2)
   xrange = np.linspace(-0.5, 0.5, 20)
   yrange = np.linspace(-0.5, 0.5, 20)
   for x in xrange:
       for y in yrange:
           if neigh.predict([x, y]) > 0:
               plt.plot(x, y, 'xr')
           else:
               plt.plot(x, y, 'xb')

   plt.title("k-NN result")
   plt.show()

This code creates the following graph:
   
.. figure:: images/xor_knn.png