我有一个数据透视表，我想创建另一个相同格式的数据透视表，但现在它包含逐年百分比变化。
这是一个简单的例子:
my_data = {
'date': [datetime.date(2000,1,7), datetime.date(2000,1,14),
datetime.date(2001,1,5), datetime.date(2001,1,12)],
'week_number': [1,2,1,2],
'quarter_number': [1,1,1,1],
'name': ['hi','bye','hi','bye'],
'category': ['clothing','electronics','clothing','electronics'],
'total sales': [123,456,180,350]
}
my_df = pd.DataFrame(my_data)
my_df.pivot_table(index=['date','week_number','quarter_number'], columns=['name', 'category'])
产生以下数据透视表:
total sales
name bye hi
category electronics clothing
date week_number quarter_number
2000-01-07 1 1 NaN 123.0
2000-01-14 2 1 456.0 NaN
2001-01-05 1 1 NaN 180.0
2001-01-12 2 1 350.0 NaN
现在假设我想计算逐年变化百分比。生成的数据透视表如下所示:
total sales pchg Y/Y
name bye hi
category electronics clothing
date week_number quarter_number
2000-01-07 1 1 NaN NaN
2000-01-14 2 1 NaN NaN
2001-01-05 1 1 NaN 0.463
2001-01-12 2 1 -0.23 NaN
请注意，在一般情况下，我们有 N 个名称、多年的数据和 K 个类别。
我在这里还提供了一个更一般的情况，以表明 pct_change 在默认模式下不起作用，因为它不会逐年发生百分比变化。
my_data = {
'date': [datetime.date(2000,1,7), datetime.date(2000,1,14),
datetime.date(2001,1,5), datetime.date(2001,1,12),
datetime.date(2000, 1, 7), datetime.date(2000, 1, 14),
datetime.date(2001, 1, 5), datetime.date(2001, 1, 12),
datetime.date(2000, 1, 7), datetime.date(2000, 1, 14),
datetime.date(2001, 1, 5), datetime.date(2001, 1, 12),
datetime.date(2000, 1, 7), datetime.date(2000, 1, 14),
datetime.date(2001, 1, 5), datetime.date(2001, 1, 12)],
'week_number': [1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2],
'quarter_number': [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1],
'name': ['hi','hi','hi','hi','hi','hi','hi','hi','bye','bye','bye','bye','bye','bye','bye','bye'],
'category': ['clothing','clothing','clothing','clothing','electronics','electronics','electronics','electronics',
'clothing', 'clothing', 'clothing', 'clothing', 'electronics', 'electronics', 'electronics','electronics'],
'total sales': [123,456,180,350,123,456,180,350,123,456,180,350,123,456,180,350]
}
my_df = pd.DataFrame(my_data)
my_df.pivot_table(index=['date','week_number','quarter_number'], columns=['name', 'category'])
my_df.pivot_table(index=['date','week_number','quarter_number'], columns=['name', 'category']).apply(pd.Series.pct_change)
total sales ...
name bye ... hi
category clothing ... electronics
date week_number quarter_number ...
2000-01-07 1 1 NaN ... NaN
2000-01-14 2 1 2.707317 ... 2.707317
2001-01-05 1 1 -0.605263 ... -0.605263
2001-01-12 2 1 0.944444 ... 0.944444
pct_change 显然是错误的，因为它不提供 Y/Y 更改，而是提供第 i 行到第 i+1 行的更改。
最佳答案
您可以使用 pct_change 获得所需的结果:
pivoted = pd.pivot_table(my_df, index=['date','week_number','quarter_number'], columns=['name', 'category'])
pivoted.groupby(level='week_number').transform(pd.Series.pct_change)
# total sales
#name bye hi
#category electronics clothing
#date week_number quarter_number
#2000-01-07 1 1 NaN NaN
#2000-01-14 2 1 NaN NaN
#2001-01-05 1 1 NaN 0.463415
#2001-01-12 2 1 -0.232456 NaN
关于python - 数据透视表中 Y 相对于 Y 的变化，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/53193333/