Question

我有一个如下数据框：

2018-10-16 15:15:00 1810.388020
2018-10-16 15:20:00 1813.299467
2018-10-16 15:25:00 1812.550495
2018-10-16 15:45:00 18140.981919
2018-10-16 15:35:00 1814.473347
2018-10-16 15:40:00 1816.349779
2018-10-16 15:45:00 740.981919
2018-10-16 15:50:00 1819.066781
2018-10-16 15:55:00 1820.333191

我需要一个如下数据框：

2018-10-16 15:15:00 1810.388020
2018-10-16 15:20:00 1813.299467
2018-10-16 15:25:00 1812.550495
2018-10-16 15:35:00 1814.473347
2018-10-16 15:40:00 1816.349779
2018-10-16 15:50:00 1819.066781
2018-10-16 15:55:00 1820.333191

即：对于每行的值大于前一行值的1.5或小于前一行值的0.5的行，将其删除

Answer 1

想法是比较前一个值和下一个值，按|链接按位OR，最后按位AND链接，按boolean indexing过滤：

m11 = df['col'] > df['col'].shift(-1) * 0.5
m12 = df['col'] > df['col'].shift() * 0.5
m21 = df['col'] < df['col'].shift(-1) * 1.5
m22 = df['col'] < df['col'].shift() * 1.5

df = df[(m11 | m12) & (m21 | m22)]
print (df)
                             col
2018-10-16 15:15:00  1810.388020
2018-10-16 15:20:00  1813.299467
2018-10-16 15:25:00  1812.550495
2018-10-16 15:35:00  1814.473347
2018-10-16 15:40:00  1816.349779
2018-10-16 15:50:00  1819.066781
2018-10-16 15:55:00  1820.333191

详细信息：

df['m11'] = df['col'] > df['col'].shift(-1) * 0.5
df['m12'] = df['col'] > df['col'].shift() * 0.5
df['m21'] = df['col'] < df['col'].shift(-1) * 1.5
df['m22'] = df['col'] < df['col'].shift() * 1.5

df['m1'] = (df['m11'] | df['m12'])
df['m2'] = (df['m21'] | df['m22'])

df['mask'] = df['m1'] & df['m2']

print (df)
                              col    m11    m12    m21    m22     m1     m2  \
2018-10-16 15:15:00   1810.388020   True  False   True  False   True   True   
2018-10-16 15:20:00   1813.299467   True   True   True   True   True   True   
2018-10-16 15:25:00   1812.550495  False   True   True   True   True   True   
2018-10-16 15:45:00  18140.981919   True   True  False  False   True  False   
2018-10-16 15:35:00   1814.473347   True  False   True   True   True   True   
2018-10-16 15:40:00   1816.349779   True   True  False   True   True   True   
2018-10-16 15:45:00    740.981919  False  False   True   True  False   True   
2018-10-16 15:50:00   1819.066781   True   True   True  False   True   True   
2018-10-16 15:55:00   1820.333191  False   True  False   True   True   True   

                      mask  
2018-10-16 15:15:00   True  
2018-10-16 15:20:00   True  
2018-10-16 15:25:00   True  
2018-10-16 15:45:00  False  
2018-10-16 15:35:00   True  
2018-10-16 15:40:00   True  
2018-10-16 15:45:00  False  
2018-10-16 15:50:00   True  
2018-10-16 15:55:00   True

根据上一行的值从数据框中删除行

1 个答案: