notebook.community

Edit and run

First class of Titanic

작성자: 박주홍

문제 정의

  • First Class 는 누가 타나요?
    • 타이타닉 데이터를 Decition Tree 에 적용
    • First Class 탑승자 Node 의 찾기
    • 위 Node 까지 도달하기 위한 Major Feature 찾기

Setting





In [1]:



    


import folium
import graphviz
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.tree import DecisionTreeClassifier, export_graphviz

pd.set_option('display.max_rows', 10)
%matplotlib inline

Read Data





In [5]:



    


titanic = sns.load_dataset('titanic')
titanic = titanic.rename(index=lambda i: f'p_{i}')
titanic



    


















    
Out[5]:











  
    

      
      survived
      pclass
      sex
      age
      sibsp
      parch
      fare
      embarked
      class
      who
      adult_male
      deck
      embark_town
      alive
      alone
    

  
  
    

      p_0
      0
      3
      male
      22.0
      1
      0
      7.2500
      S
      Third
      man
      True
      NaN
      Southampton
      no
      False
    

    

      p_1
      1
      1
      female
      38.0
      1
      0
      71.2833
      C
      First
      woman
      False
      C
      Cherbourg
      yes
      False
    

    

      p_2
      1
      3
      female
      26.0
      0
      0
      7.9250
      S
      Third
      woman
      False
      NaN
      Southampton
      yes
      True
    

    

      p_3
      1
      1
      female
      35.0
      1
      0
      53.1000
      S
      First
      woman
      False
      C
      Southampton
      yes
      False
    

    

      p_4
      0
      3
      male
      35.0
      0
      0
      8.0500
      S
      Third
      man
      True
      NaN
      Southampton
      no
      True
    

    

      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
    

    

      p_886
      0
      2
      male
      27.0
      0
      0
      13.0000
      S
      Second
      man
      True
      NaN
      Southampton
      no
      True
    

    

      p_887
      1
      1
      female
      19.0
      0
      0
      30.0000
      S
      First
      woman
      False
      B
      Southampton
      yes
      True
    

    

      p_888
      0
      3
      female
      NaN
      1
      2
      23.4500
      S
      Third
      woman
      False
      NaN
      Southampton
      no
      False
    

    

      p_889
      1
      1
      male
      26.0
      0
      0
      30.0000
      C
      First
      man
      True
      C
      Cherbourg
      yes
      True
    

    

      p_890
      0
      3
      male
      32.0
      0
      0
      7.7500
      Q
      Third
      man
      True
      NaN
      Queenstown
      no
      True
    

  



891 rows × 15 columns

Preprocessing





In [6]:



    


titanic_class = titanic[["sex", "embark_town", "age", "class"]]



    











In [7]:



    


sum(titanic_class["embark_town"].isnull())



    


















    
Out[7]:








2

















In [8]:



    


titanic_class = titanic_class[(titanic_class["embark_town"].notnull()) & (titanic_class["age"].notnull())]



    











In [9]:



    


sns.countplot(x="sex", hue="class", data=titanic_class)



    


















    
Out[9]:








<matplotlib.axes._subplots.AxesSubplot at 0x1a1d7b1f28>










    
























In [10]:



    


sns.countplot(x="embark_town", hue="class", data=titanic_class)



    


















    
Out[10]:








<matplotlib.axes._subplots.AxesSubplot at 0x110acc048>










    
























In [11]:



    


sns.catplot(x="class", y="age", kind="box", data=titanic_class)



    


















    
Out[11]:








<seaborn.axisgrid.FacetGrid at 0x110aa78d0>










    
























In [12]:



    


titanic_pre = pd.concat([titanic_class[["age", "class"]], pd.get_dummies(titanic_class[["sex", "embark_town"]])], axis=1)
titanic_pre



    


















    
Out[12]:











  
    

      
      age
      class
      sex_female
      sex_male
      embark_town_Cherbourg
      embark_town_Queenstown
      embark_town_Southampton
    

  
  
    

      p_0
      22.0
      Third
      0
      1
      0
      0
      1
    

    

      p_1
      38.0
      First
      1
      0
      1
      0
      0
    

    

      p_2
      26.0
      Third
      1
      0
      0
      0
      1
    

    

      p_3
      35.0
      First
      1
      0
      0
      0
      1
    

    

      p_4
      35.0
      Third
      0
      1
      0
      0
      1
    

    

      ...
      ...
      ...
      ...
      ...
      ...
      ...
      ...
    

    

      p_885
      39.0
      Third
      1
      0
      0
      1
      0
    

    

      p_886
      27.0
      Second
      0
      1
      0
      0
      1
    

    

      p_887
      19.0
      First
      1
      0
      0
      0
      1
    

    

      p_889
      26.0
      First
      0
      1
      1
      0
      0
    

    

      p_890
      32.0
      Third
      0
      1
      0
      1
      0
    

  



712 rows × 7 columns

Modeling





In [13]:



    


X = titanic_pre.drop("class", axis=1)
y = titanic_pre[["class"]]



    











In [14]:



    


tree_clf = DecisionTreeClassifier(max_depth=3, random_state=42)
tree_clf.fit(X, y)
print(X.columns.values)
print(tree_clf.feature_importances_)



    


















    






['age' 'sex_female' 'sex_male' 'embark_town_Cherbourg'
 'embark_town_Queenstown' 'embark_town_Southampton']
[0.66668294 0.         0.         0.33331706 0.         0.        ]

















In [15]:



    


print(np.vstack((X.columns.values, tree_clf.feature_importances_)).T)



    


















    






[['age' 0.6666829382620311]
 ['sex_female' 0.0]
 ['sex_male' 0.0]
 ['embark_town_Cherbourg' 0.333317061737969]
 ['embark_town_Queenstown' 0.0]
 ['embark_town_Southampton' 0.0]]

















In [16]:



    


export_graphviz(
        tree_clf,
        out_file="tree.dot",
        feature_names=X.columns,
        class_names=tree_clf.classes_,
        rounded=True,
        filled=True
    )



    











In [17]:



    


with open("tree.dot") as f:
    dot_graph = f.read()
dot = graphviz.Source(dot_graph)
dot.format = 'png'
dot.render(filename='tree', directory='images/decision_trees', cleanup=True)
dot



    


















    
Out[17]:













Tree



0

age <= 34.75
gini = 0.626
samples = 712
value = [184, 173, 355]
class = Third



1

embark_town_Cherbourg <= 0.5
gini = 0.558
samples = 479
value = [73, 120, 286]
class = Third



0->1


True



8

embark_town_Cherbourg <= 0.5
gini = 0.634
samples = 233
value = [111, 53, 69]
class = First



0->8


False



2

age <= 22.5
gini = 0.524
samples = 397
value = [42, 106, 249]
class = Third



1->2





5

age <= 15.5
gini = 0.624
samples = 82
value = [31, 14, 37]
class = Third



1->5





3

gini = 0.444
samples = 193
value = [19, 36, 138]
class = Third



2->3





4

gini = 0.573
samples = 204
value = [23, 70, 111]
class = Third



2->4





6

gini = 0.278
samples = 18
value = [0, 3, 15]
class = Third



5->6





7

gini = 0.618
samples = 64
value = [31, 11, 22]
class = First



5->7





9

age <= 44.5
gini = 0.662
samples = 185
value = [68, 52, 65]
class = First



8->9





12

age <= 45.75
gini = 0.19
samples = 48
value = [43, 1, 4]
class = First



8->12





10

gini = 0.647
samples = 98
value = [27, 27, 44]
class = Third



9->10





11

gini = 0.637
samples = 87
value = [41, 25, 21]
class = First



9->11





13

gini = 0.355
samples = 23
value = [18, 1, 4]
class = First



12->13





14

gini = 0.0
samples = 25
value = [25, 0, 0]
class = First



12->14























In [18]:



    


folium.Map(location=(49.634357, -1.622649), zoom_start=6)



    


















    
Out[18]:

Content source: JuhongPark/data-science

Similar notebooks:

notebook.community | gallery | about