Pandas分割DataFrame报错：'Series'对象无'startswith'属性

数据背景：info.csv结构及示例 #

device_id,upload_time,latitude,longitude,mileage,other_vals,speed,upload_time_add_8hour,upload_time_year_month,car_id,car_type,car_num,marketer_name
1234567890123,2020-09-27 02:41:02+00:00,38.01946,114.425888,0,,0,2020/9/27 10:41,202009,,,12,
17100000001,2020-09-27 02:41:01+00:00,38.01946,114.425888,0,,0,2020/9/27 10:41,202009,,,12345,
17200000002,2020-09-25 13:46:38+00:00,38.01946,114.425888,0,,0,2020/9/25 21:46,202009,,,123456,
14111111111,2020-09-25 11:18:54+00:00,38.01946,114.425888,0,,0,2020/9/25 19:18,202009,,,12121212,
57c1e18249727a0b,2020-09-25 11:18:42+00:00,38.01946,114.425888,0,,0,2020/9/25 19:18,202009,,,,
57c1e18249727a0b,2020-09-23 10:16:55+00:00,38.01946,114.425888,0.055559317,,0,2020/9/23 18:16,202009,,,,
57c1e18249727a0b,2020-09-23 10:16:15+00:00,38.01946,114.425888,0.055559317,,0,2020/9/23 18:16,202009,,,,
57c1e18249727a0b,2020-09-23 10:15:35+00:00,38.01946,114.425888,0.055559317,,0,2020/9/23 18:15,202009,,,,
57c1e18249727a0b,2020-09-23 10:15:04+00:00,38.01946,114.425888,0.055559317,,0,2020/9/23 18:15,202009,,,,
57c1e18249727a0b,2020-09-23 10:14:55+00:00,38.01946,114.425888,0.055559317,,3.304916399,2020/9/23 18:14,202009,,,,

报错代码 #

import pandas as pd
df = pd.read_csv(r'info.csv', encoding='utf-8')
df_1 = df[df['device_id'].astype(str).map(len) !=11]
df_2 = df[df['device_id'].astype(str).map(len)==11 & df['device_id'].astype(str).startswith('17')]#device_id以17开头
df_3 = df[df['device_id'].astype(str).map(len)==11 & ~df['device_id'].astype(str).startswith('17')] #device_id不以17开头
df = df[pd.notnull(df['car_num'])]
print(len(df_1))
print(len(df_2))
print(len(df_3))

问题 #

运行上述代码后出现错误：AttributeError: 'Series' object has no attribute 'startswith'，需要修复这个问题。

解决方案 #

这个报错有两个核心原因，我会逐一说明并给出修复方案：

1. 错误原因1：Series不能直接调用原生字符串方法 #

startswith是Python单个字符串的原生方法，而你这里是对Series对象调用它——Pandas的Series并没有这个属性。要对Series中的每一个字符串元素批量调用startswith，必须使用Pandas提供的str访问器，也就是str.startswith()。

2. 错误原因2：逻辑运算优先级问题 #

Python中&的优先级比==高，所以原代码里的len==11 & ...会被优先计算11 & ...，这完全违背了你的筛选逻辑。必须把每个独立的比较条件用括号括起来，再进行逻辑与运算。

修复后的完整代码 #

import pandas as pd
df = pd.read_csv(r'info.csv', encoding='utf-8')

# 修复条件的括号包裹和str访问器调用
df_1 = df[df['device_id'].astype(str).map(len) != 11]
df_2 = df[(df['device_id'].astype(str).map(len) == 11) & (df['device_id'].astype(str).str.startswith('17'))]
df_3 = df[(df['device_id'].astype(str).map(len) == 11) & (~df['device_id'].astype(str).str.startswith('17'))]

df = df[pd.notnull(df['car_num'])]
print(len(df_1))
print(len(df_2))
print(len(df_3))

额外优化建议 #

你可以把device_id转成字符串的操作提前执行一次，避免重复计算，让代码更高效整洁：

import pandas as pd
df = pd.read_csv(r'info.csv', encoding='utf-8')

# 提前转换device_id为字符串，减少重复操作
device_id_str = df['device_id'].astype(str)
df_1 = df[device_id_str.map(len) != 11]
df_2 = df[(device_id_str.map(len) == 11) & (device_id_str.str.startswith('17'))]
df_3 = df[(device_id_str.map(len) == 11) & (~device_id_str.str.startswith('17'))]

df = df[pd.notnull(df['car_num'])]
print(len(df_1))
print(len(df_2))
print(len(df_3))

这样修改后，代码就能正常运行，正确分割出三个子DataFrame并输出它们的长度了。

内容的提问来源于stack exchange，提问作者user9270170