如何在不使用.apply/.transform/.agg的前提下，基于Pandas多Groupby计算每日C/P成交量比率

Since your initial groupby gives you a multi-index Series of summed volumes per date and cp_flag, we can leverage unstacking to reshape the data into a format where we can directly compute the ratio—no need for apply/transform/agg methods.

Step-by-Step Breakdown: #

1. Get Grouped Volume Sums
   First, run your existing groupby code to get the total volume per date and cp_flag:

   grouped_volume = df.groupby(['date', 'cp_flag']).volume.sum()
   

   This gives you a Series with a multi-index (date, cp_flag) and summed volume values, like:

   date        cp_flag
   2015-01-02  C        170381
               P        366072
   2015-01-03  C        220500
               P        580000
   ...
   Name: volume, dtype: int64
   

2. Unstack to Reshape Data
   Use unstack() to pivot the cp_flag index level into columns. This turns your multi-index Series into a DataFrame where each row is a date, and columns are C and P with their respective total volumes:

   daily_cp_volumes = grouped_volume.unstack('cp_flag')
   

   Resulting DataFrame:

   cp_flag        C       P
   date                    
   2015-01-02  170381  366072
   2015-01-03  220500  580000
   ...
   

3. Compute C/P Ratio
   Now you can simply divide the C column by the P column to get your desired daily ratio as a Series:

   cp_ratio = daily_cp_volumes['C'] / daily_cp_volumes['P']
   

   The output will be exactly what you want:

   date
   2015-01-02    0.465
   2015-01-03    0.380
   ...
   2020-12-31    0.309
   dtype: float64
   

Full Combined Code: #

# Calculate grouped volume sums
grouped_volume = df.groupby(['date', 'cp_flag']).volume.sum()

# Reshape and compute ratio
daily_cp = grouped_volume.unstack('cp_flag')
cp_ratio = daily_cp['C'] / daily_cp['P']

Edge Case Note: #

If some dates are missing either C or P entries, the corresponding ratio will be NaN. If you want to handle this (e.g., fill with 0 or drop those dates), you can use dropna() or fillna() on the final cp_ratio Series:

# Drop dates with missing C/P values
cp_ratio_clean = cp_ratio.dropna()

# Or fill missing ratios with 0
cp_ratio_filled = cp_ratio.fillna(0)

内容的提问来源于stack exchange，提问作者des224