使用pandas apply时遇TypeError: string indices must be integers错误求助

错误原因 #

你调用df['列名'].apply(q1)时，apply是作用在Series的单个元素上，传入q1的参数x是列中的单个值（比如字符串'Extremely likely'或'3'），而非DataFrame的行对象。所以函数里的x['How likely are you to recommend us to a colleague?']相当于对字符串使用字符串索引，违反了Python字符串只能用整数索引的规则，导致报错。

解决方案 #

方案1：修改函数适配Series元素处理 #

调整q1函数，直接处理单个元素值：

def q1(x):
    if x == 'Extremely likely':
        return 10
    elif x == 'Not at all likely':
        return 0
    else:
        # 若原列的1-9是字符串类型，转成整数；若已是数字则直接返回
        return int(x) if isinstance(x, str) else x

调用方式不变：

df['How likely are you to recommend us to a colleague?'] = df['How likely are you to recommend us to a colleague?'].apply(q1)

方案2：用replace更简洁实现 #

不需要自定义函数，直接用字典映射替换值，再统一转类型：

mapping = {
    'Extremely likely': 10,
    'Not at all likely': 0
}
col_name = 'How likely are you to recommend us to a colleague?'
df[col_name] = df[col_name].replace(mapping).astype(int)

方案3：用numpy.where嵌套判断 #

适合逻辑简单的场景，代码更直观：

import numpy as np

col_name = 'How likely are you to recommend us to a colleague?'
df[col_name] = np.where(
    df[col_name] == 'Extremely likely',
    10,
    np.where(
        df[col_name] == 'Not at all likely',
        0,
        df[col_name].astype(int)
    )
)

内容的提问来源于stack exchange，提问作者lala345