![](/img/trans.png)
[英]Pandas DataFrame - grab value x rows below current row and compare
[英]Dataframe: compare column value and one row below
我有一个 dataframe 指示:
Direction:
2/01/19 None
1/31/19 Upward
1/30/19 None
1/29/19 None
1/28/19 Downward
1/27/19 None
1/26/19 None
1/25/19 Upward
我想根据以下条件(从 2019 年 1 月 25 日开始)创建一个“动量”列:
1. 如果对应日期的方向为“向上”,则将值设置为“向上”
2. 如果 Momentum 中下面的第一行是“向上”,则将其设置为“向上”
3. 如果对应日期的Direction为“Downward”,则设置为“None”
4. 否则,将其设置为“无”
换句话说,一旦你达到“向上”状态,它应该保持这种状态,直到你点击“向下”
结果应如下所示:
Direction: Momentum:
2/01/19 None Upward
1/31/19 Upward Upward
1/30/19 None None
1/29/19 None None
1/28/19 Downward None
1/27/19 None Upward
1/26/19 None Upward
1/25/19 Upward Upward
有没有办法在不使用循环的情况下做到这一点?
这是一种方法。 喝杯咖啡后我会尝试改进它...
df['Momentum:'] = None # Base case.
df.loc[df['Direction:'].eq('Upward'), 'Momentum:'] = 'Upward'
df.loc[df['Direction:'].eq('Downward'), 'Momentum:'] = 1 # Temporary value.
df.loc[:, 'Momentum:'] = df['Momentum:'].bfill()
df.loc[df['Momentum:'].eq(1), 'Momentum:'] = None # Set temporary value back to None.
>>> df
Direction: Momentum:
2/01/19 None Upward
1/31/19 Upward Upward
1/30/19 None None
1/29/19 None None
1/28/19 Downward None
1/27/19 None Upward
1/26/19 None Upward
1/25/19 Upward Upward
通过新数据编辑的答案首先返回填充None
值,然后将Downward
替换为None
s:
#first replace strings Nones to None type
df['Direction:'] = df['Direction:'].mask(df['Direction:'] == 'None', None)
df['Momentum:'] = df['Direction:'].bfill().mask(lambda x: x == 'Downward', None)
或者:
s = df['Direction:'].bfill()
df['Momentum:'] = s.mask(s == 'Downward', None)
print (df)
Direction: Momentum:
2/01/19 None Upward
1/31/19 Upward Upward
1/30/19 None None
1/29/19 None None
1/28/19 Downward None
1/27/19 None Upward
1/26/19 None Upward
1/25/19 Upward Upward
老答案:
使用numpy.where
与链式|
掩码比较移位值和原始值对于按位或:
mask = df['Direction:'].eq('Upward') | df['Direction:'].shift(-1).eq('Upward')
df['Momentum:'] = np.where(mask, 'Upward', None)
print (df)
Direction: Momentum:
1/31/19 None Upward
1/30/19 Upward Upward
1/29/19 None None
1/28/19 None None
1/27/19 Downward None
1/26/19 None Upward
1/25/19 Upward Upward
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.