python 数据分析实现长宽格式的转换

# encoding=utf-8

import numpy as np
import pandas as pd

# 长宽格式的转换
# 1
data = pd.read_csv('d:data/macrodata.csv')
print 'data:=\n', data
print 'data.to_records():=\n', data.to_records()
print 'data.year:=\n', data.year
print 'data.quarter:=\n', data.quarter

periods = pd.PeriodIndex(year=data.year, quarter=data.quarter, name='date')
print 'periods:=\n', periods
data = pd.DataFrame(data.to_records(),
          columns=pd.Index(['realgdp', 'infl', 'unemp'], name='item'),
          index=periods.to_timestamp('D', 'end'))
print 'data:=\n', data
ldata = data.stack().reset_index().rename(columns={0: 'value'})
# print 'ldata:=\n', ldata
print 'ldata.get(\'realgdp\'):=\n', ldata.get('realgdp')
print 'ldata.get(\'unemp\'):=\n', ldata.get('unemp')
wdata = ldata.pivot('date', 'item', 'value')
print 'ldata:=\n', ldata
print 'wdata:=\n', wdata

# 2
print 'ldata[:10]:=\n', ldata[:10]
pivoted = ldata.pivot('date', 'item', 'value')
print 'pivoted:=\n', pivoted
print 'pivoted.head():=\n', pivoted.head()
print 'ldata:=\n', ldata
ldata['value2'] = np.random.randn(len(ldata))
print 'ldata[\'value2\']:=\n', ldata['value2']
print 'ldata[:10]:=\n', ldata[:10]
pivoted = ldata.pivot('date', 'item')
print 'pivoted:=\n', pivoted
print pivoted[:5]
print 'pivoted[\'value\'][:5]:=\n', pivoted['value'][:5]
print 'ldata:=\n', ldata
unstacked = ldata.set_index(['date', 'item']).unstack('item')
print 'unstacked:=\n', unstacked

print 'test'

import pandas as pd
import numpy as np
 
 
# 常用的表格形式的数据结构
df = pd.DataFrame(np.arange(6).reshape((2,3)), index=['id1','id2'], columns=['attr1','attr2','attr3'])
print(df)
out:
  attr1 attr2 attr3
id1   0   1   2
id2   3   4   5
 
# 宽表形式(dataframe)转变为堆叠形式（key-value）形式
# 数据库中常以该形式存储
df_key_value = df.stack().reset_index()
df_key_value.columns = ['id', 'attr', 'value']
print(df_key_value)
out:
id	attr	value
0	id1	attr1	0
1	id1	attr2	1
2	id1	attr3	2
3	id2	attr1	3
4	id2	attr2	4
5	id2	attr3	5
 
# 堆叠转换为宽表形式
 
# 用set_index创建层次化索引，在用unstack重塑
# unstack中作为旋转轴的变量（如attr），其值会作为列变量展开
df_key_value.set_index(['id','attr']).unstack('attr')
out:
value
attr	attr1	attr2	attr3
id			
id1	0	1	2
id2	3	4	5
 
# 多层索引转化为宽表
df_long = df_key_value.set_index(['id','attr']).unstack('attr')['value'].reset_index()
df_long
out:
attr	id	attr1	attr2	attr3
0	id1	0	1	2
1	id2	3	4	5
 
# 堆叠转换为宽表的快捷键---pivot
df_key_value.pivot('id','attr','value')
out:
attr	attr1	attr2	attr3
id			
id1	0	1	2
id2	3	4	5