03.Regression

01.regression

 # -*- coding: utf-8 -*-
 """
 scipy 패키지 선형 회귀분석
 """
 from scipy import stats #선형 회귀분석 모듈
 import pandas as pd

 score_df=pd.read_csv("../data/score_iq.csv")
 print(score_df.info()) #150x6
 """
 RangeIndex: 150 entries, 0 to 149
 Data columns (total 6 columns):
 sid        150 non-null int64
 score      150 non-null int64
 iq         150 non-null int64
 academy    150 non-null int64
 game       150 non-null int64
 tv         150 non-null int64
 dtypes: int64(6)
 """

 print(score_df.head())
 """
      sid  score   iq  academy  game  tv
 0  10001     90  140        2     1   0
 1  10002     75  125        1     3   3
 2  10003     77  120        1     0   4
 3  10004     83  135        2     3   2
 4  10005     65  105        0     4   4
 """

 #1)단순 선현회귀분석
 #독립변수 (x:1) -> 종속변수(y:1)
 #변수 모델링
 x=score_df.iq #score_df['iq']
 y=score_df.score # #score_df['score']

 #단순 선형  회귀모형
 model=stats.linregress(x,y)

 #모델 결과
 print('model=',model)
 """
 model= LinregressResult(
 slope=0.6514309527270075, ->기울기
 intercept=-2.8564471221974657, ->절편
 rvalue=0.8822203446134699, ->설명력 1=100% 1에 가까우면 좋다
 pvalue=2.8476895206683644e-50, ->모델 유의성(0.05보다 크면 의미 없다)
 stderr=0.028577934409305443)->표준오차
 """

 #회귀방정식 =1차 함수
 #Y =aX+b (a:기울기 ,b:절편)
 #score:90   iq:140
 Y=model.slope*140-model.intercept
 print("점수 예측치=",Y) #점수 예측치= 88.34388625958358
 err=90-Y
 print("모델 오차=",err)#모델 오차= 1.6561137404164157
 print('x 기울기=',model.slope)#x 기울기= 0.6514309527270075
 print('x 절편=',model.intercept)#x 절편= -2.8564471221974657
 print('x 설명력=',model.rvalue)#x 설명력= 0.8822203446134699
 print('x 유의성=',model.pvalue)#x 유의성= 2.8476895206683644e-50
 print('x 표준오차=',model.stderr)#x 표준오차= 0.028577934409305443

 #2)다중 선형 회귀모형
 # -독립 변수 (X) 2개이상
 import statsmodels.formula.api as sm
 corr=score_df.corr()
 print("상관 계수 행렬")
 print(corr)
 """
               sid     score        iq   academy      game        tv
 sid      1.000000 -0.014399 -0.007048 -0.004398  0.018806  0.024565
 score   -0.014399  1.000000  0.882220  0.896265 -0.298193 -0.819752
 iq      -0.007048  0.882220  1.000000  0.671783 -0.031516 -0.585033
 academy -0.004398  0.896265  0.671783  1.000000 -0.351315 -0.948551
 game     0.018806 -0.298193 -0.031516 -0.351315  1.000000  0.239217
 tv       0.024565 -0.819752 -0.585033 -0.948551  0.239217  1.000000
 """

 #변수 모델 :X(iq,academy )->y(score)
 model = sm.ols(formula="score ~ iq + academy",
                data=score_df).fit()
 print("model",model) #object info
 #model <statsmodels.regression.linear_model.RegressionResultsWrapper object at 0x000000000CEAC588>

 #모델의 파라메터: 기울기 절편
 print(model.params)
 """
 Intercept    25.229141-> 절편
 iq            0.376966 ->X1 기울기
 academy       2.992800 ->X2 기울기
 dtype: float64
 """

 #다중 선형 회귀 방정식
 print(score_df.head())
 """
      sid  score   iq  academy  game  tv
 0  10001     90  140        2     1   0
 1  10002     75  125        1     3   3
 2  10003     77  120        1     0   4
 3  10004     83  135        2     3   2
 4  10005     65  105        0     4   4
 """
 Y=0.376966*140+2.992800*2+25.229141
 print("예측치=",Y)#예측치= 83.989981 

 #모델 결과
 print(model.summary())
 """
                             OLS Regression Results
 ==============================================================================
 Dep. Variable:                  score   R-squared:                       0.946
 Model:                            OLS   Adj. R-squared:                  0.946
 Method:                 Least Squares   F-statistic:                     1295.
 Date:                Sat, 16 Feb 2019   Prob (F-statistic):           4.50e-94
 Time:                        11:23:48   Log-Likelihood:                -275.05
 No. Observations:                 150   AIC:                             556.1
 Df Residuals:                     147   BIC:                             565.1
 Df Model:                           2
 Covariance Type:            nonrobust
 ==============================================================================
                  coef    std err          t      P>|t|      [0.025      0.975]
 ------------------------------------------------------------------------------
 Intercept     25.2291      2.187     11.537      0.000      20.907      29.551
 iq             0.3770      0.019     19.786      0.000       0.339       0.415
 academy        2.9928      0.140     21.444      0.000       2.717       3.269
 ==============================================================================
 Omnibus:                       36.342   Durbin-Watson:                   1.913
 Prob(Omnibus):                  0.000   Jarque-Bera (JB):               54.697
 Skew:                           1.286   Prob(JB):                     1.33e-12
 Kurtosis:                       4.461   Cond. No.                     2.18e+03
 ==============================================================================

 Warnings:
 [1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
 [2] The condition number is large, 2.18e+03. This might indicate that there are
 strong multicollinearity or other numerical problems.
 """
 """
 1.Prob (F-statistic):  4.50e-94:유의성 (0.05미만이여야 한다)
 2.Adj. R-squared:      0.946:설명력 (1에 가까와야 좋다)
 3.P>|t|                :X 유의성 검정: 0.05미만예야 좋타
 """
  

02.dot_regression

 # -*- coding: utf-8 -*-
 """
 회귀모형 예측에 행렬곱(dot) 적용예
 """
 import pandas as pd
 import numpy as np

 #1.data set 가져오기
 score_df=pd.read_csv("../data/score_iq.csv")
 print(score_df.head())# 6칼럼
 """
      sid  score   iq  academy  game  tv
 0  10001     90  140        2     1   0
 1  10002     75  125        1     3   3
 2  10003     77  120        1     0   4
 3  10004     83  135        2     3   2
 4  10005     65  105        0     4   4
 """

 #2.subset 생성
 score_arr=score_df[['score','iq','academy']]#3칼럼
 print(score_arr.shape)#(150, 3)
 print(score_arr.info())
 """
 <class 'pandas.core.frame.DataFrame'>
 RangeIndex: 150 entries, 0 to 149
 Data columns (total 3 columns):
 score      150 non-null int64
 iq         150 non-null int64
 academy    150 non-null int64
 dtypes: int64(3)
 memory usage: 3.6 KB
 None
 """
 #3.X,y변수 선택
 score_X=score_arr.ix[:,1:] #2개 (150x2) 2차원
 score_y=score_arr.ix[:,0]#1개(150) 1차원
 print(score_X.shape) #(150, 2)
 print(score_y.shape) #(150,)

 #4.기울기 ,절편
 """
 Intercept     25.229141-> 절편
 iq            0.376966 -> X1 기울기
 academy       2.992800 -> X2 기울기
 dtype: float64
 """
 #기울기 변수
 slop=np.array([[0.376966],[2.992800]]) #2차원
 Intercept=25.229141 #상수 0차원

 #Y=(a1*x1+a2*x2)+b
 #(a1*x1+a2*x2)->행렵곱

 #5.행렬곱(dot) 적용
 print(score_X.shape) #(150, 2)
 print(slop.shape) #(2, 1)
 #(150, 2) * (2, 1) =(150,1)
 matmul = np.dot(score_X,slop)
 Y = matmul + Intercept
 print(Y)
 """
 [[83.989981]
  [75.342691]
  ...
  [73.457861]]
 """

 #6. model 평가 (정답 vs 예측치)
 #Y = 예측치
 #score_y #정답
 print(Y.shape) #(150, 1)  2차원 ->1차원
 print(score_y.shape) #(150,) 1차원

 #2차원 ->1차원
 Y_fitted=Y.reshape(150) # (150,)
 df=pd.DataFrame({"Y_fitted":Y_fitted,'score':score_y})
 print(df)  # (150, 2)

 #상관 분석
 print(df.head())
 """
     Y_fitted  score
 0  83.989981     90
 1  75.342691     75
 2  73.457861     77
 3  82.105151     83
 4  64.810571     65
 """
 cor=df.Y_fitted.corr(df.score)
 print('corr=',cor) #corr= 0.9727792069594755

03.sklearn_Dataset

 # -*- coding: utf-8 -*-
 """
 sklearn 제공 datasets
 """
 from sklearn import datasets
 import numpy as np

 #1.선형회귀분석  적합 데이터셋
 #1) iris (붖꽃)
 iris=datasets.load_iris()
 print(iris)

 iris_x=iris.data #x
 iris_y=iris.target  #y

 print(type(iris_x)) #<class 'numpy.ndarray'>
 print(np.shape(iris_x)) #(150, 4)
 print(np.shape(iris_y)) #(150,)

 print(iris_x)
 """
 [[5.1 3.5 1.4 0.2]
  [4.9 3.  1.4 0.2]
  [4.7 3.2 1.3 0.2]
  [4.6 3.1 1.5 0.2]]
 """

 print(iris_y)
 """
 [0 0 ... 0 0]
 """

 #y범주
 print(list(iris.target_names)) #['setosa'=0, 'versicolor'=1, 'virginica'=2]

 #2)당뇨병 데이터셋
 diabetes=datasets.load_diabetes()
 diabetes_x=diabetes.data # x
 diabetes_y=diabetes.target # y
 print(diabetes_x.shape) #(442, 10)
 print(diabetes_y.shape) #(442,)
 print(diabetes_y)

 #3)보스톤 데이터셋
 boston=datasets.load_boston()
 boston_x=boston.data
 boston_y=boston.target
 print(boston_x.shape)#(506, 13)
 print(boston_y.shape)#(506,)
 print(boston.feature_names)
 #['CRIM' 'ZN' 'INDUS' 'CHAS' 'NOX' 'RM' 'AGE' 'DIS' 'RAD' 'TAX' 'PTRATIO' 'B' 'LSTAT']

 #2. 분류분석에 적합한 데이터셋
 #4) wine 데이터셋   다항분류 (softmax 함수)
 #'class_0:0.98,+class_1:0.01,+class_2:0.01=1
 wine= datasets.load_wine()
 wine_x=wine.data #(442, 10)
 wine_y=wine.target #(442,)
 print(wine.target_names) #['class_0' 'class_1' 'class_2']
 print(wine_x.shape)#(178, 13)
 print(wine_y)
 """
 [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
  1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
  1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
  2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2]
 """

 #5)  이진분류 (sigmoid 함수)
 # YES 0.5>  ,NO 0.5 <
 breast=datasets.load_breast_cancer()
 print(breast.data.shape) #(569, 30)
 print(breast.target.shape)#(569,)
 print(breast.target_names) #['malignant' 'benign']
 print(breast)

04.sklearn_Regression

 # -*- coding: utf-8 -*-
 """
 sklearn 관련 Regressin모델
  - y변수가 연속인 경우
 """

 import pandas as pd
 from sklearn import datasets
 from sklearn.linear_model import LinearRegression #model
 from sklearn.model_selection import train_test_split #train set VS test set
 from sklearn.metrics import mean_squared_error #MES (평균제곱 오차)

 # 1. dataset 가져오기
 iris=pd.read_csv("../data/iris.csv")
 print(iris.info())
 """
 RangeIndex: 150 entries, 0 to 149
 Data columns (total 5 columns):
 Sepal.Length    150 non-null float64
 Sepal.Width     150 non-null float64
 Petal.Length    150 non-null float64
 Petal.Width     150 non-null float64
 Species         150 non-null object
 dtypes: float64(4), object(1)
 """
 print(iris.head())
 """
    Sepal.Length  Sepal.Width  Petal.Length  Petal.Width Species
 0           5.1          3.5           1.4          0.2  setosa
 1           4.9          3.0           1.4          0.2  setosa
 2           4.7          3.2           1.3          0.2  setosa
 3           4.6          3.1           1.5          0.2  setosa
 4           5.0          3.6           1.4          0.2  setosa
 """

 #2. 변수(x,y) 선택
 cols=list(iris.columns)
 print(cols)
 #['Sepal.Length', 'Sepal.Width', 'Petal.Length', 'Petal.Width', 'Species']

 x_cols = cols[1:4] #'Sepal.Width', 'Petal.Length', 'Petal.Width'
 y_cols = cols[0] #'Sepal.Length'

 #subset
 data_df=iris[cols[:4]] #1~4칼럼
 print(data_df.shape)#(150, 4)

 #3 train set(70%)/test set(30%)   #자동 랜덤  ,random_state=123똑같은 랜덤
 iris_train,iris_test=train_test_split(
         data_df,test_size=0.3,random_state=123) 

 print(iris_train.shape)#(105, 4) model 생성
 print(iris_test.shape) #(45, 4)  model 검정

 #4.model 생성
 #help(LinearRegression)
 #class-> object
 lr_model=LinearRegression()#default model객체
 #fit(train_x,train_y) :학습->model
 lr_model.fit(iris_train[x_cols],iris_train[y_cols]) #train set

 #획귀 계수(기울기),절편
 print("기울기=",lr_model.coef_)#기울기= [ 0.63924286  0.75744562 -0.68796484]
 print("절편=",lr_model.intercept_)#절편= 1.8609363992411732

 #5. 모델 평가 :test 예측치 =회귀방정식

 #1)train set
 model_socre1=lr_model.score(iris_train[x_cols],
                             iris_train[y_cols])
 #2)test set
 model_socre2=lr_model.score(iris_test[x_cols],
                             iris_test[y_cols])

 #1.socre
 print('train_model score=',model_socre1)#train_model score= 0.8581515699458577
 print('test_model score=',model_socre2)#test_model score= 0.854680765745176

 #model 예측치 vs 정답
 pred=lr_model.predict(iris_test[x_cols])# 예측치 predict(x)
 Y=iris_test[y_cols]#정답

 #2.평균제곱오차 (MSE)
 MSE=mean_squared_error(Y,pred) #(정답,예측치)
 print('MSE=',MSE)#MSE= 0.11633863200224713

 ######################
 ### load_iris()
 ######################

 from sklearn.datasets import load_iris

 #1. data loading
 iris=load_iris()

 # 2. 변수 선택
 X=iris.data # x
 y=iris.target #y(0~2)

 print(X.shape)#(150, 4)
 print(y.shape)#(150,)

 # 3. train /test split(7:3)
 x_train,x_test,y_train,y_test=train_test_split(
         X,y, test_size=0.3,random_state=123)

 print(x_train.shape)#(105, 4) - 1~4번째
 print(x_test.shape)#(45, 4)
 print(y_train.shape)#(105,) - 5번째
 print(y_test.shape)#(45,)

 #4.model 생성:tran set
 lr_model2=LinearRegression()
 lr_model2.fit(x_train,y_train)  # train -> model

 print(lr_model2.coef_) #기울기  [-0.12591445 -0.0481559   0.24484363  0.57025678]
 print(lr_model2.intercept_) #절편 0.2537496076784179

 #5. model평가 :test set
 #1) score
 model_score=lr_model2.score(x_test,y_test)
 print(model_score) #0.9427868501294299

 #2) Mes(예측치 vs 정답)
 pred=lr_model2.predict(x_test)
 Y=y_test
 MSE=mean_squared_error(pred,Y)
 print('MSE=',MSE)#MSE= 0.04447086315865546

 #E=pred-Y
 #sqared=E^2
 import numpy as np
 mes=np.mean((pred-Y)**2)
 print('MSE=',MSE) #MSE= 0.04447086315865546

 #3시각화 평가
 import matplotlib.pyplot as plt
 fig=plt.figure(figsize=(20,5))
 chart=fig.add_subplot(1,1,1)
 chart.plot(pred,color='r',label="pred")
 chart.plot(Y,color='b',label="y")
 plt.legend(loc='best')
 plt.show()

05.LogisticRegression

 # -*- coding: utf-8 -*-
 """
 sklearn logistic Regreesion
  - y변수가 범주인 경우
 """

 from sklearn.datasets import load_iris #다항분류
 from sklearn.datasets import load_breast_cancer #이항분류
 from sklearn.linear_model import LogisticRegression

 import matplotlib.pyplot as plt
 import pandas as np

 #####################################
 ## 1. load_breast_cancer : 이항분류
 #####################################

 #1.loading data
 breast=load_breast_cancer()

 # 2. 변수 선택
 X=breast.data
 y=breast.target
 print(X.shape,y.shape)#(569, 30) (569,)

 # 3.model 생성
 #help(LogisticRegression)
 #1.random_state : 난수 seed값
 #2.solver :최적화 알고리즘
 #   {'newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga'} default: 'liblinear'
 #   작은 데이터셋:'liblinear'
 #   큰 데이터셋:'sag', 'saga'
 #   멀티 클래스 문제:'newton-cg','lbfgs'
 #   다항붕류 'multinomal'

 #적용 예)
 #1.일반 데이터셋 ,이항분류 :default
 #2일반 데이터셋 ,다항분류 :solver='lbfgs',multi_class="multinomial"
 #3.빅 데이터셋 ,이항분류 :solver='sag'

 #object
 lr_model=LogisticRegression(random_state=0)
 lr_model.fit(X,y) #model 생성 

 #예측치 predict
 pred=lr_model.predict(X)
 print('prdict=',pred[:5])#prdict= [0 0 0 1 0]
 print('y정답=',y[:5])#y정답= [0 0 0 0 0]

 # model 평가 : score = 분류정확도(accuracy)
 score=lr_model.score(X,y)
 print(score) #0.9595782073813708

 #:교차 분할표(confusing matrix)
 tab=pd.crosstab(y,pred) #crosstab(row:정답,col:예측치)
 print(tab)
 """
 col_0    0    1
 row_0
 0      198   14
 1        9  348
 """
 acc=(198+348)/len(y)
 print('accuracy=',acc)#accuracy= 0.9595782073813708

 #################################
 ## 2. load_irsi : 다항분류
 #################################
 #1.data loading
 X,y=load_iris(return_X_y=True)

 #2.model 생성
 lr_model2=LogisticRegression(random_state=123,
                              solver='lbfgs',
                              multi_class="multinomial")
 lr_model2.fit(X,y)
 print(lr_model2) #model 정보
 """
 LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,
           intercept_scaling=1, max_iter=100, multi_class='multinomial',
           n_jobs=1, penalty='l2', random_state=123, solver='lbfgs',
           tol=0.0001, verbose=0, warm_start=False)
 """
 # 예측치
 pred=lr_model2.predict(X) #예측치
 Y=y #정답

 score=lr_model2.score(X,y)
 print('accuracy=',score)#accuracy= 0.9733333333333334
 tab=pd.crosstab(Y,pred)
 print(tab)
 """
 col_0   0   1   2
 row_0
 0      50   0   0
 1       0  47   3
 2       0   1  49
 """
 print(type(tab))#<class 'pandas.core.frame.DataFrame'>

 acc=(tab.ix[0,0]+tab.ix[1,1]+tab.ix[2,2])/len(y)
 print('accuracy=',acc) #accuracy= 0.9733333333333334

 # 분류정확도(accuracy) 시각화
 import seaborn as sn # heatmap - Accuracy Score

 # confusion matrix heatmap
 plt.figure(figsize=(6,6)) # chart size
 sn.heatmap(tab, annot=True, fmt=".3f", linewidths=.5, square = True);# , cmap = 'Blues_r' : map »ö»ó
 plt.ylabel('Actual label');
 plt.xlabel('Predicted label');
 all_sample_title = 'Accuracy Score: {0}'.format(score)
 plt.title(all_sample_title, size = 18)
 plt.show()