查找 DataFrame 中两个日期期间之间是否有任何重叠

Question

我有以下pd.DataFrame

df = pd.DataFrame({'admission_timestamp': ['2021-01-17 17:45:00', '2020-03-31 23:32:00', '2020-03-27 18:20:00', '2020-04-17 18:12:00', '2020-03-19 19:12:00'], 'end_period': ['2021-01-18 17:45:00', '2020-04-01 23:32:00', '2020-03-28 18:20:00', '2020-04-18 18:12:00', '2020-03-20 19:12:00'], 'start_med': ['NaT', '2020-04-01 00:00:00', '2020-03-27 19:00:00', '2020-04-17 18:39:24', 'NaT'], 'end_med': ['NaT', '2020-04-14 21:00:00', '2020-04-05 00:00:00', '2020-05-06 22:07:29', 'NaT']})

看起来像：

   admission_timestamp  end_period           start_med            end_med 
1  2021-01-17 17:45:00  2021-01-18 17:45:00  NaT                  NaT   
2  2020-03-31 23:32:00  2020-04-01 23:32:00  2020-04-01 00:00:00  2020-04-14 21:00:00
3  2020-03-27 18:20:00  2020-03-28 18:20:00  2020-03-27 19:00:00  2020-04-05 00:00:00
4  2020-04-17 18:12:00  2020-04-18 18:12:00  2020-04-17 18:39:24  2020-05-06 22:07:29   
5  2020-03-19 19:12:00  2020-03-20 19:12:00  NaT                  NaT

我想创建一个新列 received_medidation 来说明患者是否（布尔）在 admission_timestamp 和 end_period 之间接受了药物治疗（即使只有一秒钟）。因此，布尔值应该说明admission_timestamp 和end_period 之间是否有任何时间与start_med 和end_med 之间的时间重叠。 dtypes 都是 datetime64[ns]。

我知道我们可以创建布尔掩码，例如

condition = (df['date'] > start_date) & (df['date'] <= end_date)

...但是我不明白这如何可能解决上述任务。任何帮助表示赞赏。

Answer 1

如果保证start_med 晚于admission_timestamp，那么start_med 日期在admission_timestamp 和end_period 之间就足够了

for col in df.columns:
    df[col] = pd.to_datetime(df[col])

df['received_medidation'] = (df['admission_timestamp'] < df['start_med']) & (df['start_med'] < df['end_period'])

但是，如果 start_med 可以在 admission_timestamp 之前，那么这意味着 'start_med' < 'admission_timestamp' < 'end_med' 也会创建日期的交集。然后我们使用 OR 运算符将这个案例与前一个案例包括在内：

df['received_medidation'] = (df['start_med'].between(df['admission_timestamp'], df['end_period']) |
                             df['admission_timestamp'].between(df['start_med'], df['end_med']))

注意：这里的总体假设是 admission_timestamp < end_period 和 start_med < end_med 总是正确的，在这种情况下，上面的逻辑表达式会捕获所有相交的日期。

输出：

  admission_timestamp          end_period           start_med  \
0 2021-01-17 17:45:00 2021-01-18 17:45:00                 NaT   
1 2020-03-31 23:32:00 2020-04-01 23:32:00 2020-04-01 00:00:00   
2 2020-03-27 18:20:00 2020-03-28 18:20:00 2020-03-27 19:00:00   
3 2020-04-17 18:12:00 2020-04-18 18:12:00 2020-04-17 18:39:24   
4 2020-03-19 19:12:00 2020-03-20 19:12:00                 NaT   

              end_med  received_medidation  
0                 NaT                False  
1 2020-04-14 21:00:00                 True  
2 2020-04-05 00:00:00                 True  
3 2020-05-06 22:07:29                 True  
4                 NaT                False  

Answer 2

使用between

df['overlaps'] = df['start_med'].between(df['admission_timestamp'], df['end_period']) \
                 | df['end_med'].between(df['admission_timestamp'], df['end_period'])
print(df)

# Output
  admission_timestamp          end_period           start_med             end_med  overlaps
1 2021-01-17 17:45:00 2021-01-18 17:45:00                 NaT                 NaT     False
2 2020-03-31 23:32:00 2020-04-01 23:32:00 2020-04-01 00:00:00 2020-04-14 21:00:00      True
3 2020-03-27 18:20:00 2020-03-28 18:20:00 2020-03-27 19:00:00 2020-04-05 00:00:00      True
4 2020-04-17 18:12:00 2020-04-18 18:12:00 2020-04-17 18:39:24 2020-05-06 22:07:29      True
5 2020-03-19 19:12:00 2020-03-20 19:12:00                 NaT                 NaT     False