如何用Pandas基于唯一ID的日期生成向前4天的多日期序列？

Hey there! Let's get this sorted out. The issue you're facing likely comes from two key points: your Date column is probably stored as string instead of datetime type, and your pd.date_range call was generating dates in the wrong direction. Here's how to fix it step by step:

第一步：确保日期列是datetime类型 #

First, convert your Date column to datetime format—this is critical because pd.date_range can't work directly with string dates:

df['Date'] = pd.to_datetime(df['Date'])

第二步：生成正确的回溯日期序列 #

Rewrite your function to generate 5 consecutive dates starting 4 days before the original date (inclusive). Using end=row['Date'] tells date_range to stop at the original date, and periods=5 grabs the 5 days leading up to it:

def generate_backward_dates(row):
    # Generate 5 dates ending at the original date (4 days back + original date)
    return pd.date_range(end=row['Date'], periods=5, freq='D')

df['Date_list'] = df.apply(generate_backward_dates, axis=1)

第三步：展开列表为多行 #

Use explode() to turn each list of dates into separate rows, then clean up the columns:

# Explode the date list into individual rows
result_df = df.explode('Date_list').rename(columns={'Date_list': 'Date'})
# Reset index to clean up the output
result_df = result_df.reset_index(drop=True)

完整示例代码 #

Here's the full working code with sample data:

import pandas as pd

# Sample input DataFrame
df = pd.DataFrame({
    'ID': [11, 22, 33],
    'Date': ['3/19/2018', '1/5/2018', '2/12/2018']
})

# Convert Date to datetime
df['Date'] = pd.to_datetime(df['Date'])

# Generate backward date ranges
def generate_backward_dates(row):
    return pd.date_range(end=row['Date'], periods=5, freq='D')

df['Date_list'] = df.apply(generate_backward_dates, axis=1)

# Explode and clean up
result_df = df.explode('Date_list').rename(columns={'Date_list': 'Date'}).reset_index(drop=True)

print(result_df)

This will output exactly the format you're looking for, with each ID having 5 consecutive dates starting 4 days before its original date.

内容的提问来源于stack exchange，提问作者qqqwww