pandas执行resample重采样时如何保留BG_TS等全部列

你调用resample('5T').sum()时，pandas默认仅对数值类型列执行求和聚合，BG_TS（datetime类型）、bias（字符串类型）这类非数值列会被直接丢弃，这是列丢失的核心原因。
另外你当前场景是补全5分钟间隔的时间序列，不需要对值做求和聚合，用sum()本身也不符合需求——会把同一时间窗口的SG值加总，和原始采样逻辑不符。

直接用asfreq()方法完成重采样，该方法会保留所有原始列，缺失的时间点自动填充NaN，不需要额外做0替换。同时建议在筛选子集时加.copy()，避免触发pandas的链式赋值警告。
修正后的完整代码如下：

import pandas as pd
import numpy as np

BG_test_df = pd.DataFrame(
    {'PERSON_ID': [1, 1, 1],
     'TS': ['2021-08-14 19:00:27', '2021-08-14 20:00:27', '2021-08-14 22:35:27'],             
     'bias': ["Not outside of acceptable operation. Refer to patient education","Not outside of acceptable operation. Refer to patient education","Suboptimal"]}
    )

CGM_test_df = pd.DataFrame(
    {'PERSON_ID': [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1],
     'SG': [50, 51, 52, 53, 54, 55, 400, 400, 400, 400, 400, 400, 400, 400,50, 51, 52, 53, 54, 55, 400],
     'TS': ['2021-08-14 18:30:27','2021-08-14 18:35:27','2021-08-14 18:40:27','2021-08-14 18:45:27','2021-08-14 18:50:27','2021-08-14 18:55:27', '2021-08-14 19:00:27', '2021-08-14 19:30:27','2021-08-14 19:35:27','2021-08-14 19:40:27','2021-08-14 19:45:27','2021-08-14 19:50:27','2021-08-14 19:55:27','2021-08-14 20:00:27', '2021-08-14 20:30:27','2021-08-14 20:35:27','2021-08-14 20:40:27','2021-08-14 20:45:27','2021-08-14 20:50:27','2021-08-14 20:55:27','2021-08-14 21:00:27']
     }
    )

problematic = BG_test_df.loc[BG_test_df['bias'] == "Suboptimal"].copy()

# 转换时间格式
problematic['BG_TS'] = pd.to_datetime(problematic['TS'])
CGM_test_df['CGM_TS'] = pd.to_datetime(CGM_test_df['TS'])

merged = CGM_test_df.merge(problematic, on = "PERSON_ID")

# 5分钟间隔重采样，保留所有列
filled = (merged
          .set_index('CGM_TS')
          .resample('5T')
          .asfreq()
          .reset_index()
          )

如果你后续确实需要对不同列做不同聚合计算（比如SG求和、BG_TS取窗口内第一个值），可以用agg()方法逐列指定聚合逻辑，写法如下：

filled = (merged
          .set_index('CGM_TS')
          .resample('5T')
          .agg({
              'PERSON_ID': 'first',
              'SG': 'sum',
              'TS_x': 'first', # merge后CGM侧的TS会被命名为TS_x
              'bias': 'first',
              'BG_TS': 'first',
              'TS_y': 'first' # merge后BG侧的TS会被命名为TS_y
          })
          .reset_index()
          )

内容的提问来源于stack exchange，提问作者Dr Wampa