Assignment_5_main.py

import pandas as pd
import numpy as np
from sklearn.model_selection import KFold
from sklearn.tree import DecisionTreeClassifier
from sklearn import model_selection,preprocessing, neighbors,metrics
from sklearn.utils import shuffle
from sklearn.neighbors import KNeighborsClassifier
import random
from sklearn.metrics import accuracy_score
import matplotlib.pyplot as plt
import testing,training
import os
import numpy as np


train_X =[]
test_X =[]
train_y =[]
test_y =[]
max_neighbors =15

# creating a np array for dtc-->Decision tree classifier and knn .
np_dtc = np.array([])
np_knn = np.array([])

os.mkdir('dtc')
os.mkdir('knn')
os.mkdir('accuracy')
os.mkdir('train_csv')
os.mkdir('test_csv')

data_frame = pd.read_table(r"iris.data", sep=",", header=None, names=['sepalLength', 'sepalWidth', 'petalLength', 'petalWidth', 'class'])
data_frame.head()

# Converted the last label as classes
data_frame['class'] = pd.Categorical(data_frame['class'])
data_frame['class'] = data_frame['class'].cat.codes

k_fold = model_selection.KFold(n_splits=5)


X = data_frame[data_frame.columns[:-1]].values
y = data_frame['class'].values


# Shuffling the X and y before splitting
X, y = shuffle(X, y, random_state = 999)

# Spitting the test set into 5 cross valdiation
for i, j in k_fold.split(data_frame):
    train_X.append(X[i])
    test_X.append(X[j])
    train_y.append(y[i])
    test_y.append(y[j])
    print(train_X)
    print(train_y)


for i in range(0,5):
    np.savetxt("train_csv/train_X_"+str(i)+".csv", train_X[i], delimiter=",")
    np.savetxt("train_csv/train_y_"+str(i)+".csv", train_y[i], delimiter=",")
    np.savetxt("test_csv/test_X_"+str(i)+".csv", test_X[i], delimiter=",")
    np.savetxt("test_csv/test_y_"+str(i)+".csv", test_y[i], delimiter=",")

acc_dtc = np.array([])
acc_knn = np.array([])


for i in range(5):
#     This function will train and save the model for dtc in txt format in dtc directory
    training.create_dtc(i)

#     This function will train and save the model for knn for different values of k in txt format in knn directory
    training.create_knn(i,max_neighbors)

    a = testing.run_dtc(i)
    print("Decision Tree Accuracy"+str(a))
    acc_dtc = np.append(acc_dtc,a)

    b = testing.run_knn(i,max_neighbors)
    print("KNN Accuracy"+str(b))
    acc_knn = np.append(acc_knn,b)




# reshaping in a matrix
acc_knn = acc_knn.reshape(int(5),max_neighbors)

# average mean of k= 0-max_neighbours for all 5 fold cross sets
mean_knn_accuracy = np.mean(acc_knn,axis=0)




# Mean of dtc accuracies
#  Accuracy may differ due to shuffling
np.mean(acc_dtc)




plt.bar([i+1 for i in range(0,max_neighbors)], mean_knn_accuracy)
plt.ylim([0.9,1])
plt.xlabel("Value of K")
plt.ylabel("Mean Accuracy of 5 fold")
plt.title("Mean Accuracy vs K")
plt.show()




Classifier = ('KNN', 'Decision Tree Classifier')
y_pos = np.arange(len(Classifier))
plt.bar(np.arange(2), [mean_knn_accuracy[7], np.mean(acc_dtc)])
plt.ylim([0.9,1])
plt.title("k-NN and Decision tree accuracy comparison")
plt.ylabel("Accuracy Score")
plt.xlabel("Classifiers Name")
plt.xticks(y_pos, Classifier)
plt.show()