如何用Pandas/Numpy无循环实现多维数据集行累计值重计算？

API 探路者

2026-6-17

高效实现基于前序累计和的DataFrame计算

需求说明

给定n×n的DataFrame（df_a），需生成新的n×n DataFrame（df_b），计算逻辑为：
df_b[i,j] = (df_a对应单元格的值 - 该列当前行之前所有df_b单元格的累计和) × 1×n Series（df_series）对应行的数值

示例数据与结果

n×n数据集df_a

i	0	1	2	3	4
0	0	0	0	0	0
1	0	0	0	0	0
2	200	0	0	0	0
3	200	150	0	0	0
4	200	150	125	0	0
5	200	150	125	100	0
6	200	150	125	100	75

1×n Series

i	0
0	.000
1	.001
2	.002
3	.003
4	.004
5	.005
6	.006

计算结果df_b

i	0	1	2	3	4
0	0	0	0	0	0
1	0	0	0	0	0
2	.4 [(200 - sum 0) * .002]	0	0	0	0
3	.5988 [(200 - sum .4) * .003]	.45 [(150 - sum 0) * .003]	0	0	0
4	.7960 [(200 - sum (.4, .5988)) * .004]	.5982 [(150 - sum .45) * .004]	.50	0	0
5	.9910 [(200 - sum (.4, .5988, .7960)) * .005]	.7447 [(150 - sum .45, .5982) * .005]	.6225	.50	0
6	1.1832 [(200 - sum (.4, .5988, .7960, .9910)) * .006]	.8892 [(150 - sum .45, .5982, .7447) * .006]	.74326	.5970	.450

当前实现（低效循环）

我熟练使用Pandas和Numpy，但目前通过双重循环实现该计算，耗时极长。现有循环代码如下：

import numpy as np
import pandas as pd


df_a =pd.DataFrame([[0,0,0,0,0,0,0],[0,0,0,0,0,0,0],[200,0,0,0,0,0,0],[200,150,0,0,0,0,0],[200,150,125,0,0,0,0],[200,150,125,100,0,0,0],[200,150,125,100,75,0,0]]).astype(float)

df_series = pd.Series([0.0,.001, .002, .003, .004, .005, .006])
df_b = pd.DataFrame(np.arange(49).reshape(7,7)).astype(float)

for i in range(df_a.shape[0]): # 遍历行
    for j in range(df_a.shape[1]): # 遍历列
        df_a_val = df_a.iloc[i, j] # 获取df_a对应单元格值
        ser_val = df_series.iloc[i] # 获取series对应行的值
        df_b_prev = df_b.iloc[:i,j] # 获取当前列之前所有df_b的值
        df_b_sum = df_b_prev.sum() # 计算前序值的累计和
        
        df_b.at[i, j] = (df_a_val-df_b_sum)*(ser_val) # 计算并赋值
        print("(df_a_val: {} - df_b_sum: {}) * ser_val : {} = {}".format(df_a_val, df_b_sum, ser_val, df_b.at[i, j]))
print(df_series)
print(df_b)