基于Pandas处理分时股票数据：合并9:00-9:16数据至9:16节点

数据老陈

2026-5-9

Hey there! Let's work through fixing this stock data processing problem together. The issue with your current code is that it treats all 9:00-9:16 data across every trading day as a single group, which leads to incorrect global values instead of per-day calculations. Let's fix this step by step.

Problem Breakdown

Your original code pulls all 9:00-9:16 data into one batch and uses global values (like the first open across all days, or overall max high) to update every 9:16 row. But we need to calculate these values per individual trading day since each day's pre-9:16 data is independent.

Solution Implementation

First, make sure your DataFrame's index is a datetime type (if it isn't already):

import pandas as pd

# Convert index to datetime if needed
df.index = pd.to_datetime(df.index)

Then use this code to process the data correctly:

# 1. Keep all data from 9:16 onwards (including the 9:16 timestamp itself)
post_916_data = df[df.index.time >= pd.to_datetime("09:16").time()]

# 2. Extract 9:00-9:16 data and calculate required values per trading day
pre_916_data = df.between_time("09:00", "09:16")
daily_pre_calculations = pre_916_data.groupby(pre_916_data.index.date).agg(
    open=("open", "first"),  # First open value of the 9:00-9:16 window
    high=("high", "max"),    # Highest high in the window
    low=("low", "min"),      # Lowest low in the window
    # Grab the close value exactly at 9:16
    close=("close", lambda x: x[x.index.time == pd.to_datetime("09:16").time()].iloc[0])
)

# 3. Update the 9:16 row in the post-916 data with per-day calculations
for trade_date, calc_row in daily_pre_calculations.iterrows():
    target_timestamp = pd.to_datetime(f"{trade_date} 09:16:00")
    if target_timestamp in post_916_data.index:
        post_916_data.loc[target_timestamp, ["open", "high", "low", "close"]] = calc_row.values

# 4. Final processed data (sorted to maintain time order)
final_processed_df = post_916_data.sort_index()

How This Works

Isolate post-9:16 data: We start by keeping all data from 9:16 onwards—this is our base, and we only need to modify the 9:16 rows.
Per-day pre-9:16 calculations: Grouping by trading date ensures we compute values (first open, max high, min low, 9:16 close) separately for each day.
Update 9:16 rows: We loop through each day's calculations and overwrite the corresponding 9:16 row in our base data.
Sort for consistency: Ensures the final DataFrame maintains a chronological order.

Example Verification

Using your sample data:

Original 9:16 row: 116.10 117.80 117.00 113.00
After processing, this row becomes: 116.00 117.80 116.00 113.00
Which matches your requirements: open from the first pre-9:16 entry, max high from the window, min low from the window, and close from the 9:16 timestamp.

内容的提问来源于stack exchange，提问作者amitnair92