import numpy as np
import pandas as pd
from subprocess import check_output

import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
sns.set()

print(check_output(["ls","input"]).decode('utf8'))

heart = pd.read_csv("input/heart.csv")
o2Saturation = pd.read_csv("input/o2Saturation.csv")

heart.shape, o2Saturation.shape

corr = heart.corr()
f, ax = plt.subplots(figsize=(10,8))
sns.heatmap(corr, vmax=0.8, annot=True)
abs(corr['output']).sort_values(ascending=False)

# ALL DATA TYPES are numerical data
heart.dtypes

heart.isnull().sum()

def facet_plot(feature, range_opt=None):
    facet = sns.FacetGrid(heart, hue='output', aspect=4)
    facet.map(sns.kdeplot, feature, shade = True)

    if not range_opt:
        facet.set(xlim=(0, heart[feature].max()))
    else:
        facet.set(xlim=range_opt)
    facet.add_legend()
    plt.title("Output: "+feature)

for i in heart.columns:
    if ( i != 'output'):
        facet_plot(i)

heart['cp'].value_counts()

heart['sex']=heart['sex'].apply(str)
# 사실 안해줘도 되는데, 그냥 해준거

heart.dtypes

from sklearn.preprocessing import LabelEncoder

lbl = LabelEncoder()
lbl.fit(heart['sex'])
heart['sex'] =  lbl.transform(heart['sex'].values)

heart['sex'].value_counts()

#비어있는 데이터 없음
heart.isnull().sum()

from scipy import stats
from scipy.stats import norm, skew # for some statistics

skewness = heart.apply(lambda x: skew(x.dropna()))
skewness = skewness.sort_values(ascending=False)
skewness_features = skewness[abs(skewness.values)>1].index
print("skewness:")
print(skewness_features)

sns.distplot(heart['oldpeak'], fit=norm)
(mu,sigma)=norm.fit(heart['oldpeak'])
plt.legend(['Normal dist. mu={:.2f}, std={:.2f}'.format(mu,sigma)], loc='best')
plt.ylabel('Frequency')
plt.title('Before skewness in oldpeak, skew is {:.3f}'.format(skewness['oldpeak']))
plt.show()

heart['oldpeak'] = np.log1p(heart['oldpeak'])
sns.distplot(heart['oldpeak'], fit=norm)
(mu,sigma)=norm.fit(heart['oldpeak'])
plt.legend(['Normal dist. mu={:.2f}, std={:.2f}'.format(mu,sigma)], loc='best')
plt.ylabel('Frequency')
plt.title('After skewness in oldpeak, skew is  {:.3f}'.format(skew(heart['oldpeak'])))
plt.show()

sns.distplot(heart['chol'], fit=norm)
(mu,sigma)=norm.fit(heart['chol'])
plt.legend(['Normal dist. mu={:.2f}, std={:.2f}'.format(mu,sigma)], loc='best')
plt.ylabel('Frequency')
plt.title('Before skewness in chol, skew is {:.3f}'.format(skewness['chol']))
plt.show()

heart['chol'] = np.log1p(heart['chol'])
sns.distplot(heart['chol'], fit=norm)
(mu,sigma)=norm.fit(heart['chol'])
plt.legend(['Normal dist. mu={:.2f}, std={:.2f}'.format(mu,sigma)], loc='best')
plt.ylabel('Frequency')
plt.title('After skewness in chol, skew is {:.3f}'.format(skew(heart['chol'])))
plt.show()

#QQ plots
fig = plt.figure()
res = stats.probplot(heart['chol'], plot=plt)
plt.show()

skewness = heart.apply(lambda x: skew(x.dropna()))
skewness = skewness.sort_values(ascending=False)
skewness_features = skewness[abs(skewness.values)>1].index
print("skewness:")
print(skewness_features)

label = heart['output']
heart.drop('output',axis=1, inplace=True)
heart.head()

from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score

k_fold = KFold(n_splits = 10, shuffle = True, random_state = 0)

RandomForest = make_pipeline(StandardScaler(),RandomForestClassifier(n_estimators=50) )
score = cross_val_score(RandomForest, heart, label, cv= k_fold, n_jobs =1 , scoring='accuracy')
print('RandomForest CrossValidation Score is {:.5f}'.format(np.mean(score)))

MLP = make_pipeline(StandardScaler(), MLPClassifier(learning_rate='adaptive'))
score = cross_val_score(MLP, heart, label, cv= k_fold, n_jobs =1 , scoring='accuracy')
print('MLP CrossValidation Score is {:.5f}'.format(np.mean(score)))

SVM = make_pipeline(StandardScaler(), SVC())
score = cross_val_score(SVM, heart, label, cv= k_fold, n_jobs =1 , scoring='accuracy')
print('SVM CrossValidation Score is {:.5f}'.format(np.mean(score)))

KNN = make_pipeline(StandardScaler(), KNeighborsClassifier())
score = cross_val_score(KNN, heart, label, cv= k_fold, n_jobs =1 , scoring='accuracy')
print('KNN CrossValidation Score is {:.5f}'.format(np.mean(score)))

sts = StandardScaler()
sts.fit(heart)
feed = sts.transform(heart)
feed = pd.DataFrame(feed, columns= heart.columns)

#skewness= feed.apply(lambda x: skew(x.dropna()))
#print(skewness)
#from sklearn.cluster import KMeans
#KMM = KMeans(n_clusters=2)
#KMM.fit(feed)

SVM.fit(feed, label)
pred = SVM.predict(feed)
pred = pd.DataFrame(pred, columns=['output'])

from sklearn.decomposition import PCA
pca = PCA(n_components=2)
pca.fit(feed)

X = pca.transform(feed)
X= pd.DataFrame(X)
X=X.rename(columns={0:'one', 1:'two'})
X = pd.concat([X, pred], axis=1)

colors = ['red','blue']
labels=[0,1]
legends = ['less chance of heart atk.', 'more chance of heart atk.']

for la, color, leg in zip(labels, colors, legends)W:
    plt.scatter(x=X.loc[X['output']==la]['one'], y = X.loc[X['output']==la]['two'], c =color, s=3, label=leg)
plt.legend()