[청년취업사관학교 새싹]핀테커스 수업 8주차(10/16)

<수업 내용>

연속형 data(feature)의 decision tree

import numpy as np

total_entropy=-((2/7)*np.log2(2/7)+(3/7)*np.log2(3/7)+(2/7)*np.log2(2/7))
rem_stream=(4/7)*(-((2/4)*np.log2(2/4)+(1/2)*np.log2(1/4)))+(3/7)*(-((2/3)*np.log2(2/3)+(1/3)*np.log2(1/3)))
rem_slope=(5/7)*(-((3/5)*np.log2(3/5)+(2/5)*np.log2(1/5)))

IG_stream=total_entropy-rem_stream
IG_slope=total_entropy-rem_slope

rem_treshold_750=(6/7)*(-((3/6)*np.log2(3/6)+(1/6)*np.log2(1/6)+(2/6)*np.log2(2/6)))
rem_treshold_1350=(2/7)*(-((1/2)*np.log2(1/2)+(1/2)*np.log2(1/2)))+(5/7)*(-((1/5)*np.log2(1/5)+(2/5)*np.log2(2/5)+(2/5)*np.log2(2/5)))
rem_treshold_2250=(3/7)*(-((2/3)*np.log2(2/3)+(1/3)*np.log2(1/3)))+(4/7)*(-((2/4)*np.log2(2/4)+(2/4)*np.log2(2/4)))
rem_treshold_4175=(5/7)*(-((2/5)*np.log2(2/5)+(3/5)*np.log2(3/5)))

IG_treshold_750=total_entropy-rem_treshold_750
IG_treshold_1350=total_entropy-rem_treshold_1350
IG_treshold_2250=total_entropy-rem_treshold_2250
IG_treshold_4175=total_entropy-rem_treshold_4175

print(IG_stream,IG_slope,IG_treshold_750,IG_treshold_1350,IG_treshold_2250,IG_treshold_4175)

>>
0.30595849286804166 0.5774062828523452 0.3059584928680419 0.18385092540042125 0.5916727785823274 0.863120568566631

• 첫 분기는 treshold_750가 기준이 된다

import numpy as np

total_entropy=-((2/5)*np.log2(2/5)+(3/5)*np.log2(3/5))
rem_stream=(3/5)*(-((2/3)*np.log2(2/3)+(1/3)*np.log2(1/3)))
rem_slope=(4/5)*(-((3/4)*np.log2(3/4)+(1/4)*np.log2(1/4)))
rem_treshold_750=(4/5)*(-((3/4)*np.log2(3/4)+(1/4)*np.log2(1/4)))
rem_treshold_1350=(2/5)*(-((1/2)*np.log2(1/2)+(1/2)*np.log2(1/2)))+(3/5)*(-((1/3)*np.log2(1/3)+(2/3)*np.log2(2/3)))
rem_treshold_2250=(3/5)*(-((2/3)*np.log2(2/3)+(1/3)*np.log2(1/3)))

IG_stream=total_entropy-rem_stream
IG_slope=total_entropy-rem_slope
IG_treshold_750=total_entropy-rem_treshold_750
IG_treshold_1350=total_entropy-rem_treshold_1350
IG_treshold_2250=total_entropy-rem_treshold_2250

print(IG_stream,IG_slope,IG_treshold_750,IG_treshold_1350,IG_treshold_2250)

>>
0.4199730940219749 0.3219280948873623 0.3219280948873623 0.01997309402197489 0.4199730940219749

• 두 번째 분기는 treshold_2250 혹은 IG_stream로 잡으면 된다
• IG_stream로 해보자

• 세 번째 분기가 treshold_2250된다
  

연속형 data(target)의 decision tree

1. season의 winter, spring, summer, autumn에 대한 rentals의 분산을 각각 구하면

2. 위의 값에 각각 가중치 $3/12$를 곱하여 더하면 $1379331.333$이 된다.
3. work day의 true, false에 대한 rentals의 분산을 각각 구하면

4. 위의 값에 각각 가중치 $6/12$를 곱하여 더하면 $2551813.333$이 된다
5. 분산이 작은 season을 첫 번째 분기의 기준으로 잡고, 두 번째 분기의 기준은 work day가 된다
6. leaf node의 평균을 계산하여 decision tree를 완성한다

sklearn 으로 decision tree 구현하기

from sklearn.datasets import load_iris
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split

iris=load_iris()
data, targets =iris.data, iris.target

print("data / target shape")
print(data.shape, targets.shape, '\n')

X_train, X_test, y_train, y_test=train_test_split(iris.data, iris.target, test_size=0.2, random_state=11)

print(f'{type(X_train)=}/{X_train.shape=}')
print(f'{type(X_test)=}/{X_test.shape=}')
print(f'{type(y_train)=}/{y_train.shape=}')
print(f'{type(y_test)=}/{y_test.shape=}')
#gini계수로 disicion tree그리기
model=DecisionTreeClassifier()

model.fit(X_train,y_train)

accuracy=model.score(X_test,y_test)

print(f'{accuracy=:.4f}')

import matplotlib.pyplot as plt
from sklearn import tree

plt.figure(figsize=(20,15))
tree.plot_tree(model,
               class_names=iris.target_names,
               feature_names=iris.feature_names,
               impurity=True, filled=True,
               rounded=True)

plt.show()

#entropy로 decision tree구하기
model=DecisionTreeClassifier(criterion='entropy')

model.fit(X_train,y_train)

accuracy=model.score(X_test,y_test)
import matplotlib.pyplot as plt
from sklearn import tree

plt.figure(figsize=(20,15))
tree.plot_tree(model,
               class_names=iris.target_names,
               feature_names=iris.feature_names,
               impurity=True, filled=True,
               rounded=True)

plt.show()

실습

from sklearn.datasets import load_diabetes
from sklearn.tree import DecisionTreeRegressor
from sklearn.model_selection import train_test_split
import numpy as np
diabetes=load_diabetes()
X_train, X_test, y_train, y_test=train_test_split(diabetes.data, diabetes.target, test_size=0.1, random_state=11)

model=DecisionTreeRegressor()

model.fit(X_train,y_train)

accuracy=model.score(X_test,y_test)
preds=model.predict(X_test)
>>
0.3027477160757822 
 [181.  47.  85. 128.  47. 258.  94. 272. 121. 172.  70. 272.  63. 118.
  75.  59.  48.  85. 200.  61. 153. 144. 113.  59. 164. 202. 137. 131.
 258.  88. 166. 293. 129. 118.  96.  69.  61. 110. 242. 262.  97.  91.
 233. 173. 262.]
 
mean_target=diabetes.target.mean()
denominator=np.sum((y_test-mean_target)**2)
molecule=np.sum((y_test-preds)**2)
coefficient_=1-(molecule/denominator)
coefficient_
>>
0.30431151788125865

import pandas as pd

df=pd.read_csv("./bike_sharing.csv")
from sklearn.tree import DecisionTreeRegressor
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test=train_test_split(df.iloc[:,2:-1], df['cnt'], test_size=0.2, random_state=11)

model=DecisionTreeRegressor(max_depth=3)

model.fit(X_train,y_train)

accuracy=model.score(X_test,y_test)

import matplotlib.pyplot as plt
from sklearn import tree

plt.figure(figsize=(20,15))
tree.plot_tree(model,
               class_names="cnt",
               feature_names=df.columns[:-1],
               impurity=True, filled=True,
               rounded=True)


plt.show()

import pandas as pd

df=pd.read_csv("./register_golf_club.csv")

import numpy as np

def get_entropy(df):
  
  prob_list=[]
  for j in range(len(df.columns[1:-1])):
    entry = df[f'{df.columns[1:-1][j]}'].unique()
    prob = 0

    for i in range(len(entry)):
      target=df["register_golf_club"].unique()
      
      for k in range(len(target)):
        weight = df.groupby(f"{df.columns[1:-1][j]}")["register_golf_club"].get_group(f'{entry[i]}').count()
        entropy=(df.groupby(f"{df.columns[1:-1][j]}")["register_golf_club"].get_group(f'{entry[i]}')==f'{target[k]}').sum()
        if entropy ==0:
          continue
        prob += (weight/total)*(-(entropy/weight)*np.log2(entropy/weight))
      
    prob_list.append(prob)
  return print("features별 entropy:", prob_list, "가장 작은 entropy:", min(prob_list))
  
from sklearn.preprocessing import LabelEncoder


# LabelEncoder 객체 생성
label_encoder = LabelEncoder()

# 레이블 인코딩 수행
for i in range(len(df.columns)):
  df[f'{df.columns[i]}'] = label_encoder.fit_transform(df.iloc[:,i])
  
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split


X_train, X_test, y_train, y_test=train_test_split(df.iloc[:,1:-1], df.iloc[:,-1], test_size=0.2, random_state=11)

#gini계수로 disicion tree그리기
model=DecisionTreeClassifier(criterion="entropy")

model.fit(X_train,y_train)

accuracy=model.score(X_test,y_test)

print(f'{accuracy=:.4f}')

import matplotlib.pyplot as plt
from sklearn import tree

plt.figure(figsize=(20,15))
tree.plot_tree(model,
               class_names=df.columns[1:-1],
               feature_names=df.columns[-1],
               impurity=True, filled=True,
               rounded=True)

plt.show()