Python Pandas：双条件判断与多列条件新列分类技术问题

Hey there! Let's break down your two Pandas questions with practical examples using the sample data you shared. First, here's a cleaned-up version of your sample DataFrame (fixed the incomplete product_category entry so we can actually run code against it):

import pandas as pd

df = pd.DataFrame({
    'customer_id': ['abc','abc','xyz','xyz','xyz','xyz','thr','thr','abc','abc','urt','urt'],
    'transaction_id': ['A123','A123','B345','B345','C567','C567','D678','D678','E789','E789','D903','F865'],
    'product_id': [255472, 251235, 253764,257344,221577,209809,223551,290678,908354,909238,436758,346577],
    'product_category': ['X','X','Y','Y','X','X','Z','Z','X','X','Y','Y']
})

1. Writing Multi-Condition Judgments in Pandas #

To filter or check rows based on two or more conditions, use boolean indexing with logical operators (& for AND, | for OR). Always wrap each condition in parentheses—operator precedence can trip you up otherwise!

Example 1: Filter rows where customer is 'abc' AND category is 'X' #

# AND condition: both criteria must be true
filtered_df = df[(df['customer_id'] == 'abc') & (df['product_category'] == 'X')]
print(filtered_df)

Example 2: Filter rows where transaction ID is unique OR product ID > 300000 #

# OR condition: either criteria can be true
# First, create a helper flag for unique transactions
df['is_unique_transaction'] = df['transaction_id'].duplicated(keep=False) == False

filtered_df = df[(df['is_unique_transaction']) | (df['product_id'] > 300000)]
print(filtered_df)

2. Creating a New Column with Conditional Classification #

There are a few go-to methods for this, depending on how complex your rules are:

Method 1: np.where for Simple Multi-Condition Logic #

Perfect for binary or nested binary rules. Let's create a customer_segment column with these rules:

• If customer_id is 'abc' AND product_category is 'X' → High Value
• Else if customer_id is 'xyz' AND has multiple unique transactions → Repeat Buyer
• Else → General

import numpy as np

# Pre-check if customer 'xyz' has multiple unique transactions
xyz_has_multiple_trans = df.groupby('customer_id')['transaction_id'].nunique()['xyz'] > 1

df['customer_segment'] = np.where(
    (df['customer_id'] == 'abc') & (df['product_category'] == 'X'),
    'High Value',
    np.where(
        (df['customer_id'] == 'xyz') & xyz_has_multiple_trans,
        'Repeat Buyer',
        'General'
    )
)

Method 2: df.apply() for Complex Custom Logic #

If your rules are too nested for np.where, use a custom function with apply() (note: this is slower for large DataFrames, but ideal for intricate logic):

def classify_segment(row):
    if row['customer_id'] == 'abc' and row['product_category'] == 'X':
        return 'High Value'
    elif row['customer_id'] == 'xyz':
        # Check if this customer has multiple unique transactions
        customer_trans_count = df[df['customer_id'] == row['customer_id']]['transaction_id'].nunique()
        if customer_trans_count > 1:
            return 'Repeat Buyer'
    # Catch-all for all other cases
    return 'General'

df['customer_segment'] = df.apply(classify_segment, axis=1)

Method 3: pd.cut for Numeric Range Classification #

If your classification relies on numeric ranges (like product_id), use pd.cut for clean, vectorized results:

df['product_tier'] = pd.cut(
    df['product_id'],
    bins=[0, 250000, 500000, float('inf')],
    labels=['Low Tier', 'Mid Tier', 'High Tier']
)

Pro tip: Always prioritize vectorized methods like np.where or pd.cut over apply() when possible—they’re drastically faster for large datasets!

内容的提问来源于stack exchange，提问作者jeangelj