8. 减肥对赌

  1. #  读取减肥数据集，查看一月的数据
  2.  In[63]: weight_loss = pd.read_csv('data/weight_loss.csv')
  3.          weight_loss.query('Month == "Jan"')
  4. Out[63]:

8. 减肥对赌 - 图1

  1. #  定义一个求减肥比例的函数
  2.  In[64]: def find_perc_loss(s):
  3.             return (s - s.iloc[0]) / s.iloc[0]
  4. #  查看Bob在一月的减肥成果
  5.  In[65]: bob_jan = weight_loss.query('Name=="Bob" and Month=="Jan"')
  6.          find_perc_loss(bob_jan['Weight'])
  7. Out[65]: 0    0.000000
  8.          2   -0.010309
  9.          4   -0.027491
  10.          6   -0.027491
  11.          Name: Weight, dtype: float64
  1. #  对Name和Month进行分组，然后使用transform方法，传入函数，对数值进行转换
  2.  In[66]: pcnt_loss = weight_loss.groupby(['Name', 'Month'])['Weight'].transform(find_perc_loss)
  3.          pcnt_loss.head(8)
  4. Out[66]: 0    0.000000
  5.          1    0.000000
  6.          2   -0.010309
  7.          3   -0.040609
  8.          4   -0.027491
  9.          5   -0.040609
  10.          6   -0.027491
  11.          7   -0.035533
  12.          Name: Weight, dtype: float64
  1. #  transform之后的结果，行数不变，可以赋值给原始DataFrame作为一个新列；
  2. #  为了缩短输出，只选择Bob的前两个月数据
  3.  In[67]: weight_loss['Perc Weight Loss'] = pcnt_loss.round(3)
  4.          weight_loss.query('Name=="Bob" and Month in ["Jan", "Feb"]')
  5. Out[67]:

8. 减肥对赌 - 图2

  1. #  因为最重要的是每个月的第4周，只选择第4周的数据
  2.  In[68]: week4 = weight_loss.query('Week == "Week 4"')
  3.          week4
  4. Out[68]:

8. 减肥对赌 - 图3

  1. #  用pivot重构DataFrame，让Amy和Bob的数据并排放置
  2.  In[69]: winner = week4.pivot(index='Month', columns='Name', values='Perc Weight Loss')
  3.          winner
  4. Out[69]:

8. 减肥对赌 - 图4

  1. #  用where方法选出每月的赢家
  2.  In[70]: winner['Winner'] = np.where(winner['Amy'] < winner['Bob'], 'Amy', 'Bob')
  3.          winner.style.highlight_min(axis=1)
  4. Out[70]:

8. 减肥对赌 - 图5

  1. #  用value_counts()返回最后的比分
  2.  In[71]: winner.Winner.value_counts()
  3. Out[71]: Amy    3
  4.          Bob    1
  5.          Name: Winner, dtype: int64

更多

  1. #  Pandas默认是按字母排序的
  2.  In[72]: week4a = week4.copy()
  3.          month_chron = week4a['Month'].unique() 
  4.          month_chron
  5. Out[72]: array(['Jan', 'Feb', 'Mar', 'Apr'], dtype=object)
  1. #  转换为Categorical变量，可以做成按时间排序
  2.  In[73]: week4a['Month'] = pd.Categorical(week4a['Month'], 
  3.                                           categories=month_chron,
  4.                                           ordered=True)
  5.          week4a.pivot(index='Month', columns='Name', values='Perc Weight Loss')
  6. Out[73]:

8. 减肥对赌 - 图6