회귀분석(Regression Analysis) 4

다중회귀분석(Multivariate Regression)

 

import pandas as pd

url = 'https://raw.githubusercontent.com/rusita-ai/pyData/master/Insurance.csv'
DF = pd.read_csv(url)

DF.info()

 

1. 탐색적 데이터 분석

import matplotlib.pyplot as plt
import seaborn as sns

1-1. 전체 의료비 분포

plt.figure(figsize = (9, 6))
sns.distplot(DF.expenses,
             hist = True,
             kde = True)
plt.show()

plt.figure(figsize = (9, 6))
sns.boxplot(y = 'expenses', data = DF)
plt.show()

---

1-2. 성별 별 의료비 분포

plt.figure(figsize = (9, 6))
sns.boxplot(x = 'sex', y = 'expenses', data = DF)
plt.show()

DF.sex.value_counts()

male 676

female 662

Name: sex, dtype: int64

---

1-3. 자녀수 별 의료비 분포

plt.figure(figsize = (9, 6))
sns.boxplot(x = 'children', y = 'expenses', data = DF)
plt.show()

DF.children.value_counts()

---

1-4. 흡연여부 별 의료비 분포

plt.figure(figsize = (9, 6))
sns.boxplot(x = 'smoker', y = 'expenses', data = DF)
plt.show()

DF.smoker.value_counts()

no 1064

yes 274

Name: smoker, dtype: int64

---

1-5. 거주지역 별 의료비 분포

plt.figure(figsize = (9, 6))
sns.boxplot(x = 'region', y = 'expenses', data = DF)
plt.show()

DF.region.value_counts()

---

1-6. BMI 분포 및 의료비와의 관계

# BMI 분포

plt.figure(figsize = (9, 6))
sns.distplot(DF.bmi,
             hist = True,
             kde = True)
plt.show()

# BMI와 의료비 간의 관계

plt.figure(figsize = (9, 6))
sns.scatterplot(x = DF.bmi, y = DF.expenses)
plt.show()

---

2. Modeling

2-1. Integer Encoding

• LabelEncoder( )
  • 'sex', 'smoker', 'region' to int64

from sklearn.preprocessing import LabelEncoder

encoder1 = LabelEncoder()
DF['sex'] = encoder1.fit_transform(DF.sex)

encoder2 = LabelEncoder()
DF['smoker'] = encoder2.fit_transform(DF.smoker)

encoder3 = LabelEncoder()
DF['region'] = encoder3.fit_transform(DF.region)

DF.head()

---

2-2. Train & Test Array Split(7:3)

from sklearn.model_selection import train_test_split

X = DF[['age', 'sex']]
y = DF['expenses']

X_train, X_test, y_train, y_test = train_test_split(X, y,
                                                    test_size = 0.3,
                                                    random_state = 2045)

print('Train Data : ', X_train.shape, y_train.shape)
print('Test Data : ', X_test.shape, y_test.shape)

Train Data : (936, 2) (936,)

Test Data : (402, 2) (402,)

---

2-3. fit()

from sklearn.linear_model import LinearRegression

RA = LinearRegression()
RA.fit(X_train, y_train)

LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False)

---

2-4. .predict( )

y_hat = RA.predict(X_test)

---

2-5. 오차(Error)값 확인

• Mean Sqaured Error - sklearn

from sklearn.metrics import mean_squared_error
import numpy as np

mse2 = mean_squared_error(y_test, y_hat)

np.sqrt(mse2)

10633.607635499553

---

3. Modeling

import pandas as pd

url = 'https://raw.githubusercontent.com/rusita-ai/pyData/master/Insurance.csv'
DF = pd.read_csv(url)

DF.info()

3-1. Data Preprocessing

from sklearn.model_selection import train_test_split

train_set, test_set = train_test_split(DF,
                                       test_size = 0.3,
                                       random_state = 2045)

train_set.shape, test_set.shape

((936, 7), (402, 7))

train_set.info()

---

3-2. expenses ~ age + sex

• train_set으로 모델 생성
• OLS(최소자승법) : Ordinary Least Squares

import statsmodels.formula.api as smf

Model_1 = smf.ols(formula = 'expenses ~ age + sex',
                  data = train_set).fit()

---

3-3. y_hat 생성

• test_set으로 y_hat(예측값) 계산

y_hat_1 = Model_1.predict(test_set[['age', 'sex']])

---

3-4.오차(Error)값 확인

• Mean Sqaured Error - sklearn

mse1 = mean_squared_error(test_set.expenses, y_hat_1)

np.sqrt(mse1)

10633.607635499555

---

4. Package별 오차값 비교

print('statsmodels :', np.sqrt(mse1))
print('sklearn     :', np.sqrt(mse2))

statsmodels : 10633.607635499555

sklearn : 10633.607635499553