import numpy as np # numpy 패키지 가져오기(기본패키지)
import matplotlib.pyplot as plt # 시각화 패키지 가져오기

import pandas as pd # csv -> dataframe으로 전환
from sklearn import datasets # python 저장 데이터 가져오기

from sklearn.preprocessing import StandardScaler # 연속변수의 범주화
from sklearn.preprocessing import LabelEncoder # 범주형 변수 수치화

from sklearn.model_selection import train_test_split # 훈련용과 검증용 데이터 분리

from sklearn.tree import DecisionTreeClassifier # 결정트리
from sklearn.naive_bayes import GaussianNB # 나이브 베이즈
from sklearn.neighbors import KNeighborsClassifier # K-NN(K-최근접 이웃)
from sklearn.ensemble import RandomForestClassifier # 랜덤 포레스트
from sklearn.ensemble import BaggingClassifier # 앙상블
from sklearn.linear_model import Perceptron # 퍼셉트론
from sklearn.linear_model import LogisticRegression # 로지스틱 회귀 모델
from sklearn.svm import SVC # 서포트 백터 머신(SVM)
from sklearn.neural_network import MLPClassifier #다층인공신경망

from sklearn.metrics import confusion_matrix, classification_report # 정오분류표
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, make_scorer # 정확도, 민감도
from sklearn.metrics import roc_curve # ROC 곡선 그리기

from sklearn.model_selection import cross_validate # 교차타당도
from sklearn.pipeline import make_pipeline # 파이프라인 구축
from sklearn.model_selection import learning_curve, validation_curve # 학습곡선,
from sklearn.model_selection import GridSearchCV # 9.하이퍼파라미터 튜닝

    #package load
from sklearn.preprocessing import PolynomialFeatures # 차원 조정
from sklearn.linear_model import LinearRegression    # 선형 모형
from sklearn.pipeline import make_pipeline           # 파이프라인 패키지
import numpy as np                                   # 기본 패키지

    #random date generate
np.random.seed(1)                                    # 시드 고정
X = np.random.rand(40, 1) ** 2                       # 40행, 1열
y = (10 - 1. / (X.ravel() + 0.1)) + np.random.randn(40) # np.random.randn 을 사용하여 무작위 40개 값 생성

    # plot style package
%matplotlib inline
import matplotlib.pyplot as plt
plt.style.use("seaborn-whitegrid")

    # 그림
X_test = np.linspace(-0.1, 1.1, 500).reshape(-1,1)   # -0.1에서 1.1 사이에 500개 균등하게 구간 나누고, -1,1의 2차원 reshpae

fig = plt.figure(figsize=(12, 10))
for i, degree in enumerate([1, 3, 5, 10], start=1):  # 1, 3, 5, 10 의 차원(총4개), 1부터 시작하는 인덱스
    ax = fig.add_subplot(2, 2, i)                    # 2,2,1/2,2,2/2,2,3/2,2,4
    ax.scatter(X.ravel(), y, s=15)                   # 산점도 그리기
    y_test = make_pipeline(PolynomialFeatures(degree), LinearRegression())\
             .fit(X, y).predict(X_test)              # 파이프라인 만들기
    ax.plot(X_test.ravel(), y_test, label='degree{0}'.format(degree))
    ax.set_xlim(-0.1, 1.0)
    ax.set_ylim(-2, 12)
    ax.legend(loc='best');
    
    # 검증 곡선
import numpy as np
from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import make_pipeline
import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')

np.random.seed(1)
X = np.random.rand(40, 1) ** 2
y = (10 - 1. / (X.ravel() + 0.1)) + np.random.randn(40)

from sklearn.model_selection import validation_curve
degree = np.arange(0, 21)

train_score, val_score = (make_pipeline(PolynomialFeatures(degree=2),\
                                        LinearRegression()),\
                          X, y, "polynomialfeatures__degree", degree, cv=7)

plt.figure(figsize=(8, 5))
plt.plot(degree, np.median(train_score, 1), "b-", label="training score")
plt.plot(degree, np.median(val_score, 1), "r-", label="validation score")
plt.ylim(0, 1)
plt.xlabel("degree")
plt.ylabel("score")
plt.legend(loc="best");

    # 검증 곡선에서 degree(복잡도)= 3 
X_test = np.linspace(-0.1, 1.1, 500).reshape(-1, 1)

plt.figure(figsize=(8, 7))
plt.scatter(X.ravel(), y)
lim = plt.axis()
y_pred = make_pipeline(PolynomialFeatures(degree=3), LinearRegression()).fit(X, y).predict(X_test)
plt.plot(X_test.ravel(), y_pred)
plt.axis(lim);

#실습 : overfitting & underfitting
from sklearn.preprocessing import PolynomialFeatures # 차원 조정
from sklearn.linear_model import LinearRegression    # 선형 모형
from sklearn.pipeline import make_pipeline           # 파이프라인 패키지
import numpy as np                                   # 기본 패키지

#random date generate
np.random.seed(1)                                    # 시드 고정
X = np.random.rand(40, 1) ** 2                       # 40행, 1열
y = (10 - 1. / (X.ravel() + 0.1)) + np.random.randn(40) # np.random.randn 을 사용하여 무작위 40개 값 생성

# plot style package
%matplotlib inline
import matplotlib.pyplot as plt
plt.style.use("seaborn-whitegrid")

# 그림
X_test = np.linspace(-0.1, 1.1, 500).reshape(-1,1)   # -0.1에서 1.1 사이에 500개 균등하게 구간 나누고, -1,1의 2차원 reshpae

fig = plt.figure(figsize=(12, 10))
for i, degree in enumerate([1, 3, 5, 10], start=1):  # 1, 3, 5, 10 의 차원(총4개), 1부터 시작하는 인덱스
    ax = fig.add_subplot(2, 2, i)                    # 2,2,1/2,2,2/2,2,3/2,2,4
    ax.scatter(X.ravel(), y, s=15)                   # 산점도 그리기
    y_test = make_pipeline(PolynomialFeatures(degree), LinearRegression())\
             .fit(X, y).predict(X_test)              # 파이프라인 만들기
    ax.plot(X_test.ravel(), y_test, label='degree{0}'.format(degree))
    ax.set_xlim(-0.1, 1.0)
    ax.set_ylim(-2, 12)
    ax.legend(loc='best');

# 검증 곡선
import numpy as np
from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import make_pipeline
import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')

np.random.seed(1)
X = np.random.rand(40, 1) ** 2
y = (10 - 1. / (X.ravel() + 0.1)) + np.random.randn(40)

from sklearn.model_selection import validation_curve
degree = np.arange(0, 21)

train_score, val_score = validation_curve\
(make_pipeline(PolynomialFeatures(degree=2),\
                               LinearRegression()),\
                 X, y, "polynomialfeatures__degree", degree, cv=7)

plt.figure(figsize=(8, 5))
plt.plot(degree, np.median(train_score, 1), "b-", label="training score")
plt.plot(degree, np.median(val_score, 1), "r-", label="validation score")
plt.ylim(0, 1)
plt.xlabel("degree")
plt.ylabel("score")
plt.legend(loc="best");

    # 검증 곡선에서 degree(복잡도)= 3 
X_test = np.linspace(-0.1, 1.1, 500).reshape(-1, 1)
plt.figure(figsize=(8, 7))
plt.scatter(X.ravel(), y)
lim = plt.axis()
y_pred = make_pipeline(PolynomialFeatures(degree=3), LinearRegression()).fit(X, y).predict(X_test)
plt.plot(X_test.ravel(), y_pred)
plt.axis(lim);

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier
from sklearn import preprocessing
from sklearn.model_selection import train_test_split

iris = pd.read_csv('iris.csv')
iris.head()
iris.tail()
iris.shape
iris['variety'].value_counts()
iris.columns
iris.values
iris.info() # stata의 tabulate
iris.describe() # stata의 summarize, detail
iris.describe(include='all')

X = iris.iloc[:,:4] # all row, 4th까지 column을 X에 넣기
X.head()
y = iris.iloc[:,-1] # 모든 row, 맨끝의 column을 y에 넣기
y.head()
X = preprocessing.StandardScaler().fit_transform(X)
X[0:4]

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=1) # test_size=0.3 검증을 30% 데이터로 훈련을 70% 데이터로
y_test.shape

knnmodel=KNeighborsClassifier(n_neighbors=3)
knnmodel.fit(X_train,y_train)
y_predict1=knnmodel.predict(X_test)

from sklearn.metrics import accuracy_score
acc=accuracy_score(y_test,y_predict1)
acc

from sklearn.metrics import confusion_matrix
cm=confusion_matrix(y_test.values, y_predict1)
cm

cunfmat = pd.DataFrame(confusion_matrix(y_test, y_predict1),
                        index=['True[setosa]','True[versicolor]','True[virginica]'],
                        columns=['Predict[setosa]', 'Predict[versicolor]', 'Predict[virginica]']) # 순서가 중요
cunfmat

prediction_output=pd.DataFrame(data=[y_test.values,y_predict1],index=['y_test','y_predict1'])
prediction_output.transpose()

prediction_output.iloc[0,:].value_counts()

print('잘못 분류된 샘플 개수: %d' % (y_test != y_predict1).sum()) # 뒤에 있는%를 앞에 있는 %에 넣는다는 의미
print('정확도: %.3f' % accuracy_score(y_test, y_predict1)) # 소수점 3째자리 실수로 넣기

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier
from sklearn import preprocessing
from sklearn.model_selection import train_test_split

iris = pd.read_csv('iris.csv')
iris.head()

iris.shape

iris['variety'].value_counts()

iris.columns

iris.values

iris.info() 

iris.describe()

iris.describe(include='all')

X = iris.iloc[:,:4] 
X.head()

y = iris.iloc[:,-1]
y.head()

X = preprocessing.StandardScaler().fit_transform(X)
X[0:4]

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=1)
y_test.shape

knnmodel=KNeighborsClassifier(n_neighbors=3)
knnmodel.fit(X_train,y_train)
y_predict1=knnmodel.predict(X_test)

from sklearn.metrics import accuracy_score
acc=accuracy_score(y_test,y_predict1)
acc

from sklearn.metrics import confusion_matrix
cm=confusion_matrix(y_test.values, y_predict1)
cm

cunfmat = pd.DataFrame(confusion_matrix(y_test, y_predict1),
                        index=['True[setosa]','True[versicolor]','True[virginica]'],
                        columns=['Predict[setosa]', 'Predict[versicolor]', 'Predict[virginica]']) # 순서가 중요
cunfmat

prediction_output=pd.DataFrame(data=[y_test.values,y_predict1],index=['y_test','y_predict1'])
prediction_output.transpose()

prediction_output.iloc[0,:].value_counts()

print('잘못 분류된 샘플 개수: %d' % (y_test != y_predict1).sum()) # 뒤에 있는%를 앞에 있는 %에 넣는다는 의미
print('정확도: %.3f' % accuracy_score(y_test, y_predict1)) # 소수점 3째자리 실수로 넣기