#!/usr/bin/env python
# coding: utf-8

# In[1]:


import numpy as np
import matplotlib.pyplot as plt
import sklearn.datasets as dt
import sklearn.decomposition as dec
import sklearn.neighbors as ne
import sklearn.model_selection as ms


# In[2]:


#3 features instead of 4 features:
IRIS = dt.load_iris()
X1 = IRIS.data
Y = IRIS.target

DR = dec.PCA(n_components = 3)
DR.fit(X1)
X2 = DR.transform(X1)


# In[3]:


plt.subplot(2, 2, 1)
plt.scatter(X1[:, 0], X1[:, 1], s = 7, c = Y)
plt.xlabel("X1")
plt.ylabel("X2")

plt.subplot(2, 2, 2)
plt.scatter(X2[:, 0], X2[:, 1], s = 7, c = Y)
plt.xlabel("PC1")
plt.ylabel("PC2")

plt.subplot(2, 2, 3)
plt.scatter(X1[:, 2], X1[:, 3], s = 7, c = Y)
plt.xlabel("X3")
plt.ylabel("X4")

plt.subplot(2, 2, 4)
plt.scatter(X2[:, 1], X2[:, 2], s = 7, c = Y)
plt.xlabel("PC2")
plt.ylabel("PC3")

plt.tight_layout()


# In[4]:


xtr1, xte1, ytr1, yte1 = ms.train_test_split(X1, Y, train_size = 0.75, random_state = 40)
xtr2, xte2, ytr2, yte2 = ms.train_test_split(X2, Y, train_size = 0.75, random_state = 40)


# In[5]:


KNN1 = ne.KNeighborsClassifier(n_neighbors = 5)
KNN1.fit(xtr1, ytr1)
trACC1 = KNN1.score(xtr1, ytr1)
teACC1 = KNN1.score(xte1, yte1)

KNN2 = ne.KNeighborsClassifier(n_neighbors = 5)
KNN2.fit(xtr2, ytr2)
trACC2 = KNN2.score(xtr2, ytr2)
teACC2 = KNN2.score(xte2, yte2)


# In[6]:


print("Train Accuracy Change : ", trACC2 - trACC1)
print("Test Accuracy Change : ", teACC2 - teACC1)