1、sum函数
df = pd.DataFrame([[1.4,np.nan],[7.1,-4.5],[np.nan,np.nan],[0.75,-1.3]],index = ['a','b','c','d'],columns = ['one','two']) df Out[6]: one two a 1.40 NaN b 7.10 -4.5 c NaN NaN d 0.75 -1.3 df.sum() Out[7]: one 9.25 two -5.80 dtype: float64 df.sum(axis=1) #NA值会自动排除,除非整个切片添加skipna禁用该功能 Out[8]: a 1.40 b 2.60 c 0.00 d -0.55 dtype: float64
df.mean(axis=1,skipna = False) #skipna 禁用na Out[10]: a NaN b 1.300 c NaN d -0.275 dtype: float64
df.idxmax() #idxmax和idxmin返回的是间接统计(比如达到最小值和最大值的索引) Out[13]: one b two d dtype: object df.cumsum() #案列累计 Out[14]: one two a 1.40 NaN b 8.50 -4.5 c NaN NaN d 9.25 -5.8
一次性产生多个汇总
df.describe() Out[15]: one two count 3.000000 2.000000 mean 3.083333 -2.900000 std 3.493685 2.262742 min 0.750000 -4.500000 25% 1.075000 -3.700000 50% 1.400000 -2.900000 75% 4.250000 -2.100000 max 7.100000 -1.300000
对于非数值数据,describe会产生另外一种汇总统计:
obj = pd.Series(['a','a','b','c']*4) obj.describe() Out[17]: count 16 unique 3 top a freq 8 dtype: object
2、Series的corr用于计算两个Series中重叠的、非NA的按索引对齐的相关系数,类似COV用于计算协方差
DataFrame的corr和cov将以DataFrame的形式返回完整的相关系数或协方差矩阵
利用DataFrame的corrwith方法,你可以计算其列或行跟另一个Series或DataFrame之间的相关系数。传入一个Series将会返回另一个Series(针对各列进行计算)
3、Series
value_count计算频数 obj.unique.sort()排序加唯一值
value_count适用于任何数组和序列
isin 计算一个Series各值是否包含于传入的值序列中的 布尔型数组
4、处理缺失数据
1、Series滤除缺失数据:dropna方法
DataFrame滤除用dropna方法会滤除含有NA的行和列:
from numpy import nan as NA
data = pd.DataFrame([[1.,6.5,3],[1,NA,NA],[NA,NA,NA],[NA,6.5,3.]])
cleaned = data.dropna()
data
Out[64]: 0 1 2 0 1 6.5 3 1 1 NaN NaN 2 NaN NaN NaN 3 NaN 6.5 3 cleaned Out[65]: 0 1 2 0 1 6.5 3
传入how =‘all’将只会丢弃全为NA的那些行
data.dropna(how = 'all')
Out[66]:
0 1 2 0 1 6.5 3 1 1 NaN NaN 3 NaN 6.5 3
thresh=3,会在一行中至少有 3 个非 NA 值时将其保留。
df = pd.DataFrame(np.random.randn(7,3))
df
Out[70]:
0 1 2
0 -0.182777 0.122331 1.177341
1 -0.113750 0.565018 0.460551
2 1.569241 -0.244783 -0.584233
3 1.230427 -0.749321 -0.408792
4 1.342502 1.474066 -1.078444
5 0.730472 -0.835126 0.845470
6 -0.455549 -1.051597 0.875379
df.ix[:4,1] = NA;df.ix[:2,2] = NA
df
Out[72]: 0 1 2 0 -0.182777 NaN NaN 1 -0.113750 NaN NaN 2 1.569241 NaN NaN 3 1.230427 NaN -0.408792 4 1.342502 NaN -1.078444 5 0.730472 -0.835126 0.845470 6 -0.455549 -1.051597 0.875379 df.dropna(thresh =3) Out[73]: 0 1 2 5 0.730472 -0.835126 0.845470 6 -0.455549 -1.051597 0.875379
填充缺失数据:
常数调用fillna会将缺失值替换为那个常数值:
df.fillna(0)
Out[75]:
0 1 2
0 -0.182777 0.000000 0.000000
1 -0.113750 0.000000 0.000000
2 1.569241 0.000000 0.000000
3 1.230427 0.000000 -0.408792
4 1.342502 0.000000 -1.078444
5 0.730472 -0.835126 0.845470
6 -0.455549 -1.051597 0.875379
df.fillna({1:1.5555,2:111111}) #1:1.5555代表在第一列找到NA值填充为1.5555
Out[76]: 0 1 2 0 -0.182777 1.555500 111111.000000 1 -0.113750 1.555500 111111.000000 2 1.569241 1.555500 111111.000000 3 1.230427 1.555500 -0.408792 4 1.342502 1.555500 -1.078444 5 0.730472 -0.835126 0.845470 6 -0.455549 -1.051597 0.875379
fillna默认会返回新的对象,同时也可以在旧的对象上进行修改:
_ = df.fillna(0,inplace = True)
df
Out[79]:
0 1 2
0 -0.182777 0.000000 0.000000
1 -0.113750 0.000000 0.000000
2 1.569241 0.000000 0.000000
3 1.230427 0.000000 -0.408792
4 1.342502 0.000000 -1.078444
5 0.730472 -0.835126 0.845470
6 -0.455549 -1.051597 0.875379
对reindex适用的方法同样适用于fillna:
df.fillna(method = 'ffill')
Out[82]:
0 1 2 0 0.441535 -0.330724 0.789183 1 0.666540 1.958989 -0.817482 2 -2.027360 0.416607 0.052060 3 -0.163996 -0.682905 0.926443 4 -1.163515 1.788535 -2.014879 5 0.483060 -1.225616 -0.021665 df.fillna(method='ffill',limit = 2) Out[83]: 0 1 2 0 0.441535 -0.330724 0.789183 1 0.666540 1.958989 -0.817482 2 -2.027360 0.416607 0.052060 3 -0.163996 -0.682905 0.926443 4 -1.163515 1.788535 -2.014879 5 0.483060 -1.225616 -0.021665
2、层次化索引
在一个轴上拥有多个(两个以上)的索引级别,可以低维度形式处理高纬度数据。创建一个Series,并用一个列表或数组组成的列表作为索引。
data =pd.Series(np.random.randn(10),index = [['a','a','a','b','b','b','c','c','d','d'],[1,2,3,1,2,3,1,2,2,3]]) data Out[90]: a 1 1.263309 2 -1.726591 3 0.084191 b 1 3.110524 2 -1.285000 3 0.741029 c 1 -1.443983 2 -0.775486 d 2 1.340332 3 -0.066965 dtype: float64
可用unstack聚合函数将DataFrame聚合:
data.unstack() #逆运算是stack
Out[92]:
1 2 3
a 1.263309 -1.726591 0.084191
b 3.110524 -1.285000 0.741029
c -1.443983 -0.775486 NaN d NaN 1.340332 -0.066965
重排分级顺序:
frame = pd.DataFrame(np.arange(12).reshape((4,3)),index = [['a','a','b','b'],[1,2,1,2]],columns = [['Ohio','Ohio','Colorado'],['Green','Red','Green']]) frame Out[101]: Ohio Colorado Green Red Green a 1 0 1 2 2 3 4 5 b 1 6 7 8 2 9 10 11 frame.index.names = ['key1','key2'] frame.columns.names = ['state','color'] frame Out[104]: state Ohio Colorado color Green Red Green key1 key2 a 1 0 1 2 2 3 4 5 b 1 6 7 8 2 9 10 11
用swaplevel来进行转换外层级别:
frame.swaplevel('key1','key2') Out[105]: state Ohio Colorado color Green Red Green key2 key1 1 a 0 1 2 2 a 3 4 5 1 b 6 7 8 2 b 9 10 11
sortlevel(1) swaplevel(0,1) 对索level进行排序
使用DataFrame的列:DataFrame的sex_index函数会将其一个或多个列转换为行索引,并创建一个新索引
frame = pd.DataFrame({'a':range(7),'b':range(7,0,-1),'c':['one','one','one','two','two','two','tow'],'d':[0,1,2,0,1,2,3]}) frame Out[115]: a b c d 0 0 7 one 0 1 1 6 one 1 2 2 5 one 2 3 3 4 two 0 4 4 3 two 1 5 5 2 two 2 6 6 1 tow 3 frame2 = frame.set_index(['c','d']) #默认情况下,那些列会从DataFrame中移除,但可以添加 drop = False 来 frame2 Out[117]: a b c d one 0 0 7 1 1 6 2 2 5 two 0 3 4 1 4 3 2 5 2 tow 3 6 1
不考虑索引类型的、基于位置的索引,可以使用Series的iget_value方法和DataFrame的irow和icol方法:
ser3 = pd.Series(range(3),index = [-5,1,3])
ser3.iget_value(2) #基于索引位置的值
Out[129]: 2
frame = pd.DataFrame(np.arange(6).reshape(3,2),index = [2,0,1])