我想计算数据集中服务器停止的时间长度。 我知道停机时间，但不知道持续时间。
我有这个 df:
index a b c reboot stop
2018-06-25 12:49:00 NaN NaN NaN 0 1
2018-06-25 12:50:00 NaN NaN NaN 0 1
2018-06-25 12:51:00 NaN NaN NaN 1 1
2018-06-25 12:52:00 NaN NaN NaN 0 1
2018-06-25 12:53:00 NaN NaN NaN 0 1
2018-06-25 12:54:00 NaN NaN NaN 0 1
2018-06-25 12:55:00 NaN NaN NaN 0 1
2018-06-25 12:56:00 NaN NaN 1.2 0 0
2018-06-25 12:57:00 NaN NaN NaN 0 1
2018-06-25 12:58:00 NaN NaN NaN 1 1
2018-06-25 12:59:00 NaN NaN NaN 0 1
2018-06-25 13:00:00 NaN NaN NaN 0 1
2018-06-25 13:01:00 NaN NaN NaN 0 0
如果a、b、c = NaN，
我的服务器在 reboot, stop = 1 时停止
并在reboot, stop = 0时启动。
期望的输出:
index period
2018-06-25 12:51:00 5
2018-06-25 12:58:00 3
最佳答案
这将实现你想要的:
# Create a new column which identifies stopped times
df['stopped'] = np.nan
idx_stopped = (pd.isnull(df.a)) & (pd.isnull(df.b)) & (pd.isnull(df.c)) & (df.reboot == 1) & (df.stop == 1)
df.loc[idx_stopped, 'stopped'] = 1
df.loc[(df.reboot == 0) & (df.stop == 0), 'stopped'] = 0
df.stopped = df.stopped.ffill()
df.stopped = df.stopped.fillna(0)
df.loc[df.stopped == 0, 'stopped'] = np.nan
# Count the number of periods for each stop instance
v = df.stopped[::-1]
cumsum = v.cumsum().fillna(method='pad')
reset = -cumsum[v.isnull()].diff().fillna(cumsum)
result = v.where(v.notnull(), reset).cumsum()
df['period'] = result[::-1]
# Identify the time each stop incident began
df['first'] = (df.stopped == 1) & (pd.isnull(df.stopped.shift(1)))
df2 = df[['index', 'period']][df['first']]
df2.period = df2.period.astype(int)
print(df2)
index period
2 2018-06-25 12:51:00 5
9 2018-06-25 12:58:00 3
关于python - 数据框 pandas 中的条件(NaN 列)和条件，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/55462233/