如何在Pandas中按每日06:00~08:00时间范围筛选多重索引数据？

Hey there! Let's figure out how to filter your MultiIndex DataFrame to only keep rows between 6:00 and 8:00 every day. I've got two solid approaches for you:

方法1：直接针对Time层级的索引筛选 #

Since your DataFrame uses Date and Time as a MultiIndex, we can directly target the Time level to filter our desired range. This keeps your original index structure intact:

import pandas as pd

# Optional: If your 'Time' index is stored as strings, convert it to datetime.time type first for safer comparisons
time_level = pd.to_datetime(df.index.get_level_values('Time')).time
df.index = df.index.set_levels(time_level, level='Time')

# Define our target time range
start = pd.to_datetime('06:00:00').time()
end = pd.to_datetime('08:00:00').time()

# Filter the DataFrame
filtered_df = df[(df.index.get_level_values('Time') >= start) & 
                 (df.index.get_level_values('Time') <= end)]

If your Time values are already in datetime.time format, you can skip the conversion step and jump straight to filtering.

方法2：合并索引为Datetime后使用between_time #

If you prefer a more intuitive approach (and might need to do more time-based operations later), merging the Date and Time levels into a single datetime index makes things super easy with pandas' built-in between_time method:

import pandas as pd

# Combine Date and Time into a single datetime column
df['full_datetime'] = pd.to_datetime(
    df.index.get_level_values('Date').astype(str) + ' ' + 
    df.index.get_level_values('Time').astype(str)
)

# Set this new column as the index
df = df.set_index('full_datetime')

# Filter rows between 6:00 and 8:00 daily
filtered_df = df.between_time('06:00', '08:00')

# Optional: Restore the original MultiIndex if needed
filtered_df['Date'] = filtered_df.index.date
filtered_df['Time'] = filtered_df.index.time
filtered_df = filtered_df.set_index(['Date', 'Time'])

Quick Notes #

• If your Time values are strings like 5:15:00 (without leading zeros), converting to datetime.time is still safe—pandas handles that automatically.
• The first method is great if you want to keep your original index structure without modifications. The second method is more flexible for future time-based analysis.

内容的提问来源于stack exchange，提问作者jerry han