7. 使用Seaborn网格做多变量分析

  1. #  读取employee数据集，创建工龄的列
  2.  In[85]: employee = pd.read_csv('data/employee.csv', 
  3.                                 parse_dates=['HIRE_DATE', 'JOB_DATE'])
  4.          days_hired = (pd.to_datetime('12-1-2016') - employee['HIRE_DATE'])
  6.          one_year = pd.Timedelta(1, unit='Y')
  7.          employee['YEARS_EXPERIENCE'] =  days_hired / one_year
  8.          employee[['HIRE_DATE', 'YEARS_EXPERIENCE']].head()
  9. Out[85]:

7. 使用Seaborn网格做多变量分析 - 图1

  1. #  画一个基本的带有回归线的散点图
  2.  In[86]: import seaborn as sns
  3.  In[87]: ax = sns.regplot(x='YEARS_EXPERIENCE', y='BASE_SALARY',
  4.                           data=employee)
  5.          ax.figure.set_size_inches(14,4)
  6. Out[87]:

7. 使用Seaborn网格做多变量分析 - 图2

  1. #  用regplot的上层函数lmplot，画出不同性别的回归线
  2.  In[88]: grid = sns.lmplot(x='YEARS_EXPERIENCE', y='BASE_SALARY',
  3.                            hue='GENDER', palette='Greys',
  4.                            scatter_kws={'s':10}, data=employee)
  5.          grid.fig.set_size_inches(14, 4)
  6.          type(grid)
  7. Out[88]: seaborn.axisgrid.FacetGrid

7. 使用Seaborn网格做多变量分析 - 图3

  1. #  为每个种族创建子图，同时保留回归线
  2.  In[89]: grid = sns.lmplot(x='YEARS_EXPERIENCE', y='BASE_SALARY',
  3.                            hue='GENDER', col='RACE', col_wrap=3,
  4.                            palette='Greys', sharex=False,
  5.                            line_kws = {'linewidth':5},
  6.                            data=employee)
  7.          grid.set(ylim=(20000, 120000))
  8. Out[89]: <seaborn.axisgrid.FacetGrid at 0x11e7ce470>

7. 使用Seaborn网格做多变量分析 - 图4

  1. #  将类型值的层级减小到二，将部门的层级减小到三
  2.  In[90]: deps = employee['DEPARTMENT'].value_counts().index[:2]
  3.          races = employee['RACE'].value_counts().index[:3]
  4.          is_dep = employee['DEPARTMENT'].isin(deps)
  5.          is_race = employee['RACE'].isin(races)
  6.          emp2 = employee[is_dep & is_race].copy()
  7.          emp2['DEPARTMENT'] = emp2.DEPARTMENT.str.extract('(HPD|HFD)', expand=True)
  8.          emp2.shape
  9. Out[90]: (968, 11)
  11.  In[91]: emp2['DEPARTMENT'].value_counts()
  12. Out[91]: HPD    591
  13.          HFD    377
  14.          Name: DEPARTMENT, dtype: int64
  16.  In[92]: emp2['RACE'].value_counts()
  17. Out[92]: White                        478
  18.          Hispanic/Latino              250
  19.          Black or African American    240
  20.          Name: RACE, dtype: int64
  1. #  用Axe层函数，比如violinplot来画出工龄和性别的分布
  2.  In[93]: ax = sns.violinplot(x = 'YEARS_EXPERIENCE', y='GENDER', data=emp2)
  3.          ax.figure.set_size_inches(10,4)
  4. Out[93]:

7. 使用Seaborn网格做多变量分析 - 图5

  1. #  用factorplot函数，为每个部门和种族的组合画图
  2.  In[94]: sns.factorplot(x ='YEARS_EXPERIENCE', y='GENDER',
  3.                         col='RACE', row='DEPARTMENT', 
  4.                         size=3, aspect=2,
  5.                         data=emp2, kind='violin')
  6. Out[94]: <seaborn.axisgrid.FacetGrid at 0x11e40ec50>

7. 使用Seaborn网格做多变量分析 - 图6