目录
1.axis与skipna参数的使用
注意：np.nan表示空值
# 建立数据集
import numpy as np
import pandas as pd
df = pd.DataFrame({\'key1\':[4,5,6,np.nan,2], # np.nan表示空值
\'key2\':[1,2,np.nan,15,10],
\'key3\':[1,2,3,\'m\',\'n\'],
\'key4\':[\'a1\',\'a2\',\'a3\',\'a4\',\'a5\']},index = [\'a\',\'b\',\'c\',\'d\',\'e\']
)
print(df)
print(\'-----------------分割线-------------------\')
# 打印每列数据类型
print(df[\'key1\'].dtype,df[\'key2\'].dtype,df[\'key3\'].dtype,df[\'key4\'].dtype)
# 求整体平均值（只会对数字列进行统计）
print(\'----------------整体平均------------------\')
print(df.mean())
# 根据索引求平均值
print(\'key2列的平均值：\',df[\'key2\'].mean())
print(\'----------------axis参数------------------\')
# 单独统计一列的平均值,axis=0表示以列计算 axis=1表示以行计算，默认axis=0
m1 = df.mean(axis=1)
print(m1)
print(\'----------------skipna参数------------------\')
# skipna参数：是否忽略NaN，默认True，如False，有NaN的列统计结果仍未NaN
m2 = df.mean(skipna=False)
print(m2)
输出结果：
2.常用统计方法
其他随机方法列表如下：
- count：统计非Na值的数量
- min:统计最小值
- max:统计最大值
- quantile:统计分位数，参数q确定位置,例如：quantile(q=0.75)
- sum:求和
- median:求中位数
- std:标准差
- var：方差
- skew:偏度
- kurt:峰度
实战演练：
import numpy as np
import pandas as pd
df = pd.DataFrame({\'key1\':[1,2,3,4,25,6,7,8,9,10],
\'key2\':[10,20,30,300,50,60,70,80,90,100]})
print(df)
print(\'-\' * 20)
# 求最大值
print("计数count：")
print(df.count())
print(\'-\' * 20)
print("最小值min：")
print(df.min())
print(\'-\' * 20)
print("最大值min：")
print(df.max())
print(\'-\' * 20)
print("第三分位数quantile：")
print(df.quantile(0.75))
print(\'-\' * 20)
print("求和sum：")
print(df.sum())
print(\'-\' * 20)
print("求中位数median：")
print(df.median())
print(\'-\' * 20)
print("求标准差std：")
print(df.std())
print(\'-\' * 20)
print("求方差var：")
print(df.var())
print(\'-\' * 20)
print("求偏度skew：")
print(df.skew())
print(\'-\' * 20)
print("求峰度kurt：")
print(df.kurt())
输出结果：
3.求累计值:cumsum(累计和),累计积
import numpy as np
import pandas as pd
df = pd.DataFrame({\'key1\':[1,2,3,4,25,6,7,8,9,10],
\'key2\':[10,20,30,300,50,60,70,80,90,100]})
print(df)
print(\'-------------累计和&累计积-----------------\')
df[\'key1累计和\'] = df[\'key1\'].cumsum()
df[\'key2累计和\'] = df[\'key2\'].cumsum()
df[\'key1累计积\'] = df[\'key1\'].cumsum()
df[\'key2累计积\'] = df[\'key1\'].cumsum()
print(df)
输出结果：
4.唯一值：unique
import pandas as pd
s = pd.Series(list(\'abdsssdsd\'))
print(s)
# 求唯一值
sq = s.unique()
print(\'-\' * 50)
print(sq,type(sq))
输出结果：
5.值计数：value_counts()
值计数主要针对新的Series，计算出不同值出现的频率
import pandas as pd
s = pd.Series(list(\'abdsssdsd\'))
print(s)
print(\'-\' * 50)
print(s.value_counts(sort = False)) # sort参数：排序，默认为True
输出结果：
6.成员资格：isin()
判断某个数据集的元素是否在另外一个数据集中，返回结果为布尔值。
s = pd.Series(np.arange(10,13))
df = pd.DataFrame({\'key1\':list(\'asdc\'),
\'key2\':np.arange(4,8)})
print(s)
print(\'-\' * 50)
print(df)
print(\'-\' * 50)
print(s.isin([12,15]))
print(\'-\' * 50)
print(df.isin([\'a\',\'bc\',\'10\',8]))
输出结果：