Pandas 创建层次化索引

1.创建多层次索引

1.1 隐式构造

• 最常见的方法是给DataFrame构造函数的index参数传递两个或更多的数组

# 导入pandasimport numpy as npimport pandas as pd​data = np.random.randint(0,100,size=(6,6))​# 行索引index = [    ["1班","1班","1班","2班","2班","2班"],    ["张三","李四","王五","鲁班","张三丰","张无忌"]]​# 列索引columns = [    ["期中","期中","期中","期末","期末","期末"],    ["语文","数学","英语","语文","数学","英语"]]​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	40	51	30	16	57	45
	李四	74	45	70	48	75	1
	王五	57	45	35	25	22	76
2班	鲁班	93	80	69	31	17	29
	张三丰	90	38	36	77	56	30
	张无忌	35	50	79	45	38	76

• Series也可以创建多层索引

data = np.random.randint(0,100,size=6)​index = [    ["1班","1班","1班","2班","2班","2班"],    ["张三","李四","王五","鲁班","张三丰","张无忌"]]​s = pd.Series(data=data,index=index)s

1班  张三      7
    李四      9
    王五     57
2班  鲁班     88
    张三丰    36
    张无忌     5
dtype: int32

1.2 显式构造pd.MultiIndex

• 使用数组

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_arrays([    ["1班","1班","1班","2班","2班","2班"],    ["张三","李四","王五","鲁班","张三丰","张无忌"]])​# 列索引columns = [    ["期中","期中","期中","期末","期末","期末"],    ["语文","数学","英语","语文","数学","英语"]]​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	56	0	70	16	54	65
	李四	11	99	94	66	82	51
	王五	37	16	71	4	82	72
2班	鲁班	39	33	65	69	77	68
	张三丰	53	15	23	99	79	7
	张无忌	53	30	18	95	7	36

• 使用tuple

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_tuples(    (        ("1班","张三"),("1班","李四"),("1班","王五"),        ("2班","鲁班"),("2班","张三丰"),("2班","张无忌")    ))​# 列索引columns = [    ["期中","期中","期中","期末","期末","期末"],    ["语文","数学","英语","语文","数学","英语"]]​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	27	89	20	7	5	1
	李四	75	60	58	79	50	15
	王五	37	47	56	43	59	55
2班	鲁班	41	25	43	71	46	37
	张三丰	66	53	52	21	53	91
	张无忌	29	26	22	49	56	24

• 使用product

笛卡尔积：{a,b}{c,d} ==> {a,c},{a,d},{b,c},{b,d}

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_product([    ["1班","2班"],    ["张三","李四","王五"]])​# 列索引columns = [    ["期中","期中","期中","期末","期末","期末"],    ["语文","数学","英语","语文","数学","英语"]]​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	75	54	83	5	27	56
	李四	52	63	1	10	63	84
	王五	67	6	5	96	45	16
2班	张三	77	10	10	94	41	73
	李四	86	34	51	50	18	87
	王五	9	73	91	43	38	45

2.多层列索引

除了行索引index，列索引columns也能用同样的方法创建多层索引

• 使用数组

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_arrays([    ["1班","1班","1班","2班","2班","2班"],    ["张三","李四","王五","鲁班","张三丰","张无忌"]])​# 列索引columns = pd.MultiIndex.from_arrays([    ["期中","期中","期中","期末","期末","期末"],    ["语文","数学","英语","语文","数学","英语"]])​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	20	58	72	66	62	71
	李四	67	22	63	46	16	21
	王五	34	92	15	1	74	30
2班	鲁班	18	97	24	5	50	86
	张三丰	58	24	17	32	49	52
	张无忌	50	33	26	38	41	82

• 使用tuple

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_tuples(    (        ("1班","张三"),("1班","李四"),("1班","王五"),        ("2班","鲁班"),("2班","张三丰"),("2班","张无忌")    ))​# 列索引columns = pd.MultiIndex.from_tuples(    (        ("期中","语文"),("期中","数学"),("期中","英语"),        ("期末","语文"),("期末","数学"),("期末","英语")    ))​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	55	37	66	70	6	4
	李四	38	32	79	79	78	2
	王五	80	49	56	51	32	19
2班	鲁班	36	68	81	13	35	73
	张三丰	94	56	94	45	15	34
	张无忌	78	5	93	57	24	65

• 使用product

data = np.random.randint(0,100,size=(6,6))​# 行索引index = pd.MultiIndex.from_product([    ["1班","2班"],    ["张三","李四","王五"]])​# 列索引columns = pd.MultiIndex.from_product([    ["期中","期末"],    ["语文","数学","英语"]])​df = pd.DataFrame(data=data,index=index,columns=columns)df

		期中			期末
		语文	数学	英语	语文	数学	英语
1班	张三	1	18	17	4	94	72
	李四	41	33	22	75	36	77
	王五	42	82	28	21	84	57
2班	张三	18	64	9	0	41	80
	李四	24	99	66	92	34	25
	王五	28	41	16	2	85	36