import pandas as pd
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn import preprocessing
from sklearn.neighbors import  KNeighborsClassifier
from sklearn.decomposition import PCA
import matplotlib.pyplot as plt

import sklearn.metrics as met

def classification(message, x_arg, y_arg):

    print(message, '\n')

    # podela na trening i test skup
    x_train, x_test, y_train, y_test = train_test_split(x_arg, y_arg, train_size=0.7, stratify=y)

    # Parametri za unakrsnu validacuju
    parameters = [{'n_neighbors': range(3, 15, 2),
                   'p': [1, 2],
                   'weights': ['uniform', 'distance']
                   }]

    clf = GridSearchCV(KNeighborsClassifier(), parameters, cv=10)
    clf.fit(x_train, y_train)

    print("Najbolji parametri:")
    print(clf.best_params_)
    print()

    print("Izvestaj za trening skup:")
    y_pred =clf.predict(x_train)
    print(met.classification_report(y_train, y_pred))
    print()

    cnf_matrix = met.confusion_matrix(y_train, y_pred)
    print("Matrica konfuzije", cnf_matrix, sep="\n")
    print("\n")

    print("Izvestaj za test skup:")
    y_pred =clf.predict(x_test)
    print(met.classification_report(y_test, y_pred))
    print()

    cnf_matrix = met.confusion_matrix(y_test, y_pred)
    print("Matrica konfuzije", cnf_matrix, sep="\n")
    print("\n")

    if message == 'PCA':
        colors = ['red', 'blue', 'gold',  'm', 'plum', 'orange', 'black']
        x_test.is_copy=False
        x_test['predicted'] = y_pred
        classes =x_test['predicted'].unique()
        for i, class_value in zip(range(0, len(classes)), classes):
            class_samples = x_test.loc[lambda s: s['predicted'] == class_value, :]
            plt.scatter(class_samples['pca1'], class_samples['pca2'], color=colors[i],
                        s=10, marker='o', label="class %s" % class_value)

        plt.title('Classification with PCA')
        plt.legend(loc='upper right')
        plt.show()


df = pd.read_csv("car.csv")

#prikaz imena kolona + 5 prvih instanci
print('Prvih 5 instanci', df.head(), sep='\n')
print('\n\n')


print('Opis podataka', df.describe(), sep='\n')
print('\n\n')

print('Klase:', print(df["class"].value_counts()), sep='\n')
print('\n\n')

features=df.columns[1:]
x=df[features]
y=df["class"]


num_features = x.shape[1]

#standardizacija podataka
scaler = preprocessing.StandardScaler().fit(x)
x =pd.DataFrame(scaler.transform(x))
x.columns = features

#primena pca
pca=PCA()
#pca=PCA(n_components=2)
pca.fit(x)
x_pca = pd.DataFrame(pca.transform(x))

#promena imena kolona za skup sa pca
pca_columns = ['pca%d'%i for i in range(1, pca.n_components_+1)]
x_pca.columns=pca_columns

print('components_ ')
for i, component in zip(range(1, pca.n_components_+1), pca.components_):
    pca_desc="pca%d"%i + "="
    for j, value in zip(range(0, num_features), component):
        pca_desc+="%.2f*%s"%(value, features[j])
    print(pca_desc)

print()

print('explained_variance_  ')
for i, ev in zip(range(1, num_features+1), pca.explained_variance_):
    print("pca%d: %.10f"%(i,ev))
print()

print()

print('explained_variance_ratio_  ')

for i, evr in zip(range(1, num_features+1), pca.explained_variance_ratio_):
    print("pca%d: %.10f"%(i,evr))
print()


print('mean_  ', pca.mean_ , sep='\n')
print()


print('n_components_  ', pca.n_components_ , sep='\n')
print()

print('noise_variance_   ', pca.noise_variance_  , sep='\n')
print()

classification('Original', x, y)
classification('PCA', x_pca, y)