Python Pandas筛选数据的30种方法，效率翻倍

# 筛选年龄大于30的记录
df[df['age'] > 30]

# 筛选年龄大于30且工资低于50000的记录
df[(df['age'] > 30) & (df['salary'] < 50000)]

# 使用字符串表达式筛选
df.query("age > 25 and department == 'Sales'")

# 筛选姓名包含"John"的记录
df[df['name'].str.contains('John', na=False)]

# 筛选邮箱以"gmail.com"结尾的记录
df[df['email'].str.endswith('gmail.com')]

# 筛选符合电话号码格式的记录
df[df['phone'].str.match(r'^\d{3}-\d{3}-\d{4}$')]

# 筛选年龄在25到35之间的记录
df[df['age'].between(25, 35)]

# 筛选工资在前20%的记录
df[df['salary'] > df['salary'].quantile(0.8)]

# 筛选温度变化绝对值大于5的记录
df[df['temperature_change'].abs() > 5]

# 筛选部门为销售或市场的记录
df[df['department'].isin(['Sales', 'Marketing'])]

# 筛选部门不是HR或IT的记录
df[~df['department'].isin(['HR', 'IT'])]

# 筛选2024年的记录
df[df['date'].between('2024-01-01', '2024-12-31')]

# 筛选3月份的记录
df[df['date'].dt.month == 3]

# 筛选周末的记录
df[df['date'].dt.dayofweek.isin([5, 6])]

# 筛选邮箱不为空的记录
df[df['email'].notna()]

# 筛选电话号码为空的记录
df[df['phone'].isna()]

# 筛选数值型列
numeric_cols = df.select_dtypes(include=['number'])

# 排除对象类型列
non_object_cols = df.select_dtypes(exclude=['object'])

# 筛选特定行和列
df.loc[df['score'] > 90, ['name', 'score']]

# 筛选前100行的第2,4,6列
df.iloc[:100, [1, 3, 5]]

# 筛选每个部门工资前3的员工
df.groupby('department').apply(lambda x: x.nlargest(3, 'salary'))

# 筛选成员数超过10的部门
df.groupby('department').filter(lambda x: len(x) > 10)

# 筛选工资最高的5条记录
df.nlargest(5, 'salary')

# 筛选最早入职的10名员工
df.sort_values('hire_date').head(10)

# 随机抽取50条记录
df.sample(n=50, random_state=42)

# 随机抽取10%的记录
df.sample(frac=0.1, random_state=42)

# 筛选姓名和邮箱重复的记录
df[df.duplicated(subset=['name', 'email'], keep=False)]

# 去除重复记录，保留第一次出现
df.drop_duplicates(subset=['name', 'email'], keep='first')

# 使用自定义函数筛选复杂条件
df[df.apply(lambda row: len(row['name']) > 5 and row['age'] < 40, axis=1)]

# 对于大型数据集使用eval提升性能
df[df.eval('age > 30 and salary > 50000')]