Python循环生成time_of_day列异常：全列仅显示Morning求助

问题场景 #

尝试通过for循环+if语句给Pandas DataFrame新增time_of_day列，规则是：

• 6≤Hour≤12 → Morning
• 12<Hour≤17 → Afternoon
• 17<Hour≤22 → Evening
• 其余 → Night

但执行代码后，time_of_day列全是Morning，value_counts()只统计到1428条Morning。

错误代码 #

for i in range(len(df)):
    if df.loc[i, 'Hour'] >=6 & df.loc[i, 'Hour'] <= 12:
        df['time_of_day'] = "Morning"

    elif df.loc[i, 'Hour'] > 12 & df.loc[i, 'Hour'] <= 17:
        df['time_of_day'] == "Afternoon"
        
    elif df.loc[i, 'Hour'] > 17 & df.loc[i, 'Hour'] <= 22:
        df['time_of_day'] == "Evening"
    
    else:
        df['time_of_day'] = 'Night' 

df.time_of_day.value_counts()   

错误原因分析 #

1. 整列赋值而非单行赋值：每次循环直接给df['time_of_day']赋值，会覆盖整个列的所有值。比如最后一次循环如果满足Morning的条件，整列就会被改成Morning，之前的赋值全部无效。应该用df.loc[i, 'time_of_day']指定单行赋值。
2. 逻辑运算符错误：Python中逻辑与要用and，&是位运算符，且未加括号时优先级会导致判断逻辑出错。正确写法应为(df.loc[i, 'Hour'] >=6) and (df.loc[i, 'Hour'] <=12)。
3. 赋值运算符用错：elif分支里用了==（比较判断）而非=（赋值），等于根本没给列赋值，即使条件满足也不会改变列的值。

修正后的for循环代码 #

# 先初始化列，避免赋值时出现缺失
df['time_of_day'] = ''

for i in range(len(df)):
    hour = df.loc[i, 'Hour']
    if 6 <= hour <= 12:
        df.loc[i, 'time_of_day'] = "Morning"
    elif 12 < hour <= 17:
        df.loc[i, 'time_of_day'] = "Afternoon"
    elif 17 < hour <= 22:
        df.loc[i, 'time_of_day'] = "Evening"
    else:
        df.loc[i, 'time_of_day'] = 'Night'

df.time_of_day.value_counts()

更高效的Pandas原生实现（推荐） #

Pandas不推荐用for循环处理数据，效率太低，用pd.cut或者np.select更简洁高效：

方法1：pd.cut

bins = [-float('inf'), 6, 12, 17, 22, float('inf')]
labels = ['Night', 'Morning', 'Afternoon', 'Evening', 'Night']
df['time_of_day'] = pd.cut(df['Hour'], bins=bins, labels=labels, include_lowest=True)

df.time_of_day.value_counts()

方法2：np.select

import numpy as np

conditions = [
    (df['Hour'] >=6) & (df['Hour'] <=12),
    (df['Hour'] >12) & (df['Hour'] <=17),
    (df['Hour'] >17) & (df['Hour'] <=22)
]
choices = ['Morning', 'Afternoon', 'Evening']
df['time_of_day'] = np.select(conditions, choices, default='Night')

df.time_of_day.value_counts()

内容的提问来源于stack exchange，提问作者Sami