KnnModel.py

# coding: utf-8
from collections import Counter
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import accuracy_score

class KNearestNeightbors(object):
  def __init__(self, k = 1):
      self._train_data = None
      self._target_data = None
      self._k = k
      
  def fit(self, train_data, target_data):
      self._train_data = train_data
      self._target_data = target_data
  
  def predict(self, x):
      distances = np.array([self._distance(p, x) for p in self._train_data])
      nearest_indexes = distances.argsort()[:self._k]
      nearest_labels = self._target_data[nearest_indexes]
      c = Counter(nearest_labels)
      
      return c.most_common(1)[0][0]
  
  def _distance(self, p0, p1):
      return np.sum((p0 - p1) ** 2)
  
  def find_x2(self, cls0, cls1, x1):
      diff = []
      x2 = np.linspace(0.5, 3.0, 100)
      for y in x2:
          x = [x1, y]
          distances_0 = np.array([self._distance(p, x) for p in cls0])
          distances_1 = np.array([self._distance(p, x) for p in cls1])
          diff.append(np.absolute(np.min(np.array(distances_0)) - np.min(np.array(distances_1))))
      x2_indexes = np.argmin(diff)
      return x2[x2_indexes]

def main():
  iris_dataset = np.loadtxt("dataset/iris_traning.csv", delimiter=",")
  features = iris_dataset[:, 1:3]
  targets = iris_dataset[:, 0]
  iris_dataset_test = np.loadtxt("dataset/iris_test.csv", delimiter=",")
  features_test = iris_dataset_test[:, 1:3]
  targets_test = iris_dataset_test[:, 0]

  for k in [1, 5, 10]:
      model = KNearestNeightbors(k)
      model.fit(features, targets)
      predicted_labels = []
      for test in features_test:
          predicted_label = model.predict(test)
          predicted_labels.append(predicted_label)
      score = accuracy_score(targets_test, predicted_labels)
      print("k = {}".format(k) )
      print("acuracy : {}".format(score))

  cls0 = iris_dataset[targets == 0, 1:3]
  cls1 = iris_dataset[targets > 0, 1:3]
  x1 = np.linspace(3, 7, 100)
  x2 = [model.find_x2(cls0, cls1, x) for x in x1]

  plt.scatter(features[:,0], features[:,1], linewidths=0, alpha=1,
              c=targets
             )
  plt.plot(x1, x2)
  plt.show()

if __name__ == '__main__':
  main()