groupby 的妙用(注意size和count)
Pandas的groupby()
功能很强大,用好了可以方便的解决很多问题,在数据处理以及日常工作中经常能施展拳脚。
今天,我们一起来领略下groupby()
的魅力吧。
首先,引入相关package:
import pandas as pd
import numpy as np
groupby的基础操作
df = pd.DataFrame({'A': ['a', 'b', 'a', 'c', 'a', 'c', 'b', 'c'], ...: 'B': [2, 8, 1, 4, 3, 2, 5, 9], ...: 'C': [102, 98, 107, 104, 115, 87, 92, 123]}) ...: df ...: Out[2]: A B C 0 a 2 102 1 b 8 98 2 a 1 107 3 c 4 104 4 a 3 115 5 c 2 87 6 b 5 92 7 c 9 123
按A列分组(groupby),获取其他列的均值
df.groupby('A').mean() Out[3]: B C A a 2.0 108.000000 b 6.5 95.000000 c 5.0 104.666667
按多列进行分组(groupby)
df.groupby(['A','B']).mean() Out[4]: C A B a 1 107 2 102 3 115 b 5 92 8 98 c 2 87 4 104 9 123
分组后选择列进行运算
In [5]: df = pd.DataFrame([[1, 1, 2], [1, 2, 3], [2, 3, 4]], columns=["A", "B", "C"]) ...: df ...: Out[5]: A B C 0 1 1 2 1 1 2 3 2 2 3 4 In [6]: g = df.groupby("A") In [7]: g['B'].mean() # 仅选择B列 Out[7]: A 1 1.5 2 3.0 Name: B, dtype: float64 In [8]: g[['B', 'C']].mean() # 选择B、C列 Out[8]: B C A 1 1.5 2.5 2 3.0 4.0
可以针对不同的列选用不同的聚合方法
In [9]: g.agg({'B':'mean', 'C':'sum'}) Out[9]: B C A 1 1.5 5 2 3.0 4
聚合方法size()和count()
size跟count的区别: size计数时包含NaN值,而count不包含NaN值
In [10]: df = pd.DataFrame({"Name":["Alice", "Bob", "Mallory", "Mallory", "Bob" , "Mallory"], ...: "City":["Seattle", "Seattle", "Portland", "Seattle", "Seattle", "Portland"], ...: "Val":[4,3,3,np.nan,np.nan,4]}) ...: ...: df ...: Out[10]: City Name Val 0 Seattle Alice 4.0 1 Seattle Bob 3.0 2 Portland Mallory 3.0 3 Seattle Mallory NaN 4 Seattle Bob NaN 5 Portland Mallory 4.0
count()
In [11]: df.groupby(["Name", "City"], as_index=False)['Val'].count() Out[11]: Name City Val 0 Alice Seattle 1 1 Bob Seattle 1 2 Mallory Portland 2 3 Mallory Seattle 0
size()
In [12]: df.groupby(["Name", "City"])['Val'].size().reset_index(name='Size') Out[12]: Name City Size 0 Alice Seattle 1 1 Bob Seattle 2 2 Mallory Portland 2 3 Mallory Seattle 1
分组运算方法 agg()
In [13]: df = pd.DataFrame({'A': list('XYZXYZXYZX'), 'B': [1, 2, 1, 3, 1, 2, 3, 3, 1, 2], ...: 'C': [12, 14, 11, 12, 13, 14, 16, 12, 10, 19]}) ...: df ...: Out[13]: A B C 0 X 1 12 1 Y 2 14 2 Z 1 11 3 X 3 12 4 Y 1 13 5 Z 2 14 6 X 3 16 7 Y 3 12 8 Z 1 10 9 X 2 19 In [14]: df.groupby('A')['B'].agg({'mean':np.mean, 'standard deviation': np.std}) Out[14]: mean standard deviation A X 2.250000 0.957427 Y 2.000000 1.000000 Z 1.333333 0.577350
针对不同的列应用多种不同的统计方法
In [15]: df.groupby('A').agg({'B':[np.mean, 'sum'], 'C':['count',np.std]}) Out[15]: B C mean sum count std A X 2.250000 9 4 3.403430 Y 2.000000 6 3 1.000000 Z 1.333333 4 3 2.081666
分组运算方法 apply()
In [16]: df = pd.DataFrame({'A': list('XYZXYZXYZX'), 'B': [1, 2, 1, 3, 1, 2, 3, 3, 1, 2], ...: 'C': [12, 14, 11, 12, 13, 14, 16, 12, 10, 19]}) ...: df ...: Out[16]: A B C 0 X 1 12 1 Y 2 14 2 Z 1 11 3 X 3 12 4 Y 1 13 5 Z 2 14 6 X 3 16 7 Y 3 12 8 Z 1 10 9 X 2 19 In [17]: df.groupby('A').apply(np.mean) ...: # 跟下面的方法的运行结果是一致的 ...: # df.groupby('A').mean() Out[17]: B C A X 2.250000 14.750000 Y 2.000000 13.000000 Z 1.333333 11.666667
apply()
方法可以应用lambda函数,举例如下:
In [18]: df.groupby('A').apply(lambda x: x['C']-x['B']) Out[18]: A X 0 11 3 9 6 13 9 17 Y 1 12 4 12 7 9 Z 2 10 5 12 8 9 dtype: int64 In [19]: df.groupby('A').apply(lambda x: (x['C']-x['B']).mean()) Out[19]: A X 12.500000 Y 11.000000 Z 10.333333 dtype: float64
分组运算方法 transform()
前面进行聚合运算的时候,得到的结果是一个以分组名为 index 的结果对象。如果我们想使用原数组的 index 的话,就需要进行 merge 转换。transform(func, args, *kwargs) 方法简化了这个过程,它会把 func 参数应用到所有分组,然后把结果放置到原数组的 index 上(如果结果是一个标量,就进行广播):
In [20]: df = pd.DataFrame({'group1' : ['A', 'A', 'A', 'A', ...: 'B', 'B', 'B', 'B'], ...: 'group2' : ['C', 'C', 'C', 'D', ...: 'E', 'E', 'F', 'F'], ...: 'B' : ['one', np.NaN, np.NaN, np.NaN, ...: np.NaN, 'two', np.NaN, np.NaN], ...: 'C' : [np.NaN, 1, np.NaN, np.NaN, ...: np.NaN, np.NaN, np.NaN, 4]}) ...: df ...: Out[20]: B C group1 group2 0 one NaN A C 1 NaN 1.0 A C 2 NaN NaN A C 3 NaN NaN A D 4 NaN NaN B E 5 two NaN B E 6 NaN NaN B F 7 NaN 4.0 B F In [21]: df.groupby(['group1', 'group2'])['B'].transform('count') Out[21]: 0 1 1 1 2 1 3 0 4 1 5 1 6 0 7 0 Name: B, dtype: int64 In [22]: df['count_B']=df.groupby(['group1', 'group2'])['B'].transform('count') ...: df ...: Out[22]: B C group1 group2 count_B 0 one NaN A C 1 1 NaN 1.0 A C 1 2 NaN NaN A C 1 3 NaN NaN A D 0 4 NaN NaN B E 1 5 two NaN B E 1 6 NaN NaN B F 0 7 NaN 4.0 B F 0
上面运算的结果分析: {‘group1’:’A’, ‘group2’:’C’}的组合共出现3次,即index为0,1,2。对应”B”列的值分别是”one”,”NaN”,”NaN”,由于count()计数时不包括Nan值,因此{‘group1’:’A’, ‘group2’:’C’}的count计数值为1。
transform()方法会将该计数值在dataframe中所有涉及的rows都显示出来(我理解应该就进行广播)
将某列数据按数据值分成不同范围段进行分组(groupby)运算
In [23]: np.random.seed(0) ...: df = pd.DataFrame({'Age': np.random.randint(20, 70, 100), ...: 'Sex': np.random.choice(['Male', 'Female'], 100), ...: 'number_of_foo': np.random.randint(1, 20, 100)}) ...: df.head() ...: Out[23]: Age Sex number_of_foo 0 64 Female 14 1 67 Female 14 2 20 Female 12 3 23 Male 17 4 23 Female 15
打铁还需自身硬