如何对Pandas中groupby分组后的每组应用自定义得分计算函数？

错误点分析 #

1. 字符串匹配错误：原始数据中DPP列的 promoter 值是Promoter，但你定义的函数里写成了Promoters（多了末尾的s），导致无法正确统计数量。
2. 调用语法错误：df.groupby('Cust_type').apply(score, x))存在两处问题：一是apply会自动将每个分组的子DataFrame传入score函数，不需要手动传递x参数；二是末尾多了一个括号，触发语法报错。

修正后的完整代码 #

import pandas as pd

# 构造原始DataFrame
data = {
    'Id': [1,2,3,4,5,6,7,8,9,10,11,12],
    'Cust_type': ['A','B','C','D','A','B','B','C','D','D','A','A'],
    'DPP': ['Detractor','Detractor','Detractor','Promoter','Promoter','Passive','Detractor','Detractor','Detractor','Promoter','Promoter','Passive']
}
df = pd.DataFrame(data)

# 修正后的得分计算函数
def score(x):
    promoter_num = len(x[x['DPP'] == 'Promoter'])
    detractor_num = len(x[x['DPP'] == 'Detractor'])
    total = len(x)
    return ((promoter_num - detractor_num) / total) * 100

# 正确调用groupby.apply
grouped_score = df.groupby('Cust_type').apply(score)
print(grouped_score)

运行后输出：

Cust_type
A    25.000000
B   -66.666667
C  -100.000000
D     0.000000
dtype: float64

更高效的优化写法 #

如果处理大数据量，推荐用value_counts来统计各类别数量，避免多次子DataFrame切片，提升性能：

def score_optimized(x):
    count_dict = x['DPP'].value_counts().to_dict()
    promoter = count_dict.get('Promoter', 0)
    detractor = count_dict.get('Detractor', 0)
    total = sum(count_dict.values())
    return ((promoter - detractor) / total) * 100

grouped_score_optimized = df.groupby('Cust_type').apply(score_optimized)
print(grouped_score_optimized)

输出结果和之前一致，但运行效率更高。

内容的提问来源于stack exchange，提问作者lala345