如何在pandas DataFrame中根据指定值获取列索引？

Hey there! Let's work through how to find the column index for the value "Mobile : 0" in your DataFrame. You mentioned iloc isn't working here—and that makes sense, since iloc is designed for position-based access, not searching for specific values. Here are a couple of reliable approaches to solve this:

Method 1: Boolean Filtering + Column Index Lookup #

This method uses boolean matching to identify columns containing your target value, then converts that column label to its integer index:

import pandas as pd
import numpy as np

# Your input data converted to DataFrame
data = [['25362438,25383532 Mobile : 8691017781,8691017798', np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, 'Mobile : 0', np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan]]
df = pd.DataFrame(data)

target_value = "Mobile : 0"

# Step 1: Find columns that contain the target value
matching_columns = df.columns[df.eq(target_value).any()]

# Step 2: Convert the column label to its integer index
column_index = df.columns.get_loc(matching_columns[0])

print(column_index)  # Output: 16

How this works: #

• df.eq(target_value) creates a boolean DataFrame where each cell is True if it matches your target value, otherwise False.
• .any() checks each column to see if there's at least one True (i.e., the column contains the target value).
• df.columns.get_loc() converts the column label (in your case, an integer like 16) to its corresponding position—this is especially handy if your columns have non-integer names later on.

Method 2: Flatten Data with stack() #

Another approach is to flatten the DataFrame into a Series (automatically removing NaN values), then extract the column index from the resulting multi-index:

# Flatten the DataFrame to a Series (drops NaNs automatically)
stacked_data = df.stack()

# Find the row where the value matches, then get the column index
matching_entry = stacked_data[stacked_data == target_value].index[0]
column_index = matching_entry[1]  # The second element of the index tuple is the column

print(column_index)  # Output: 16

How this works: #

• stack() pivots column labels into the row index, creating a multi-index (row number, column number) for each non-NaN value.
• We filter the Series to find entries matching the target, then pull the column index directly from the multi-index tuple.

Why iloc isn't the right tool here #

iloc requires you to already know the position of the row/column you want to access. Since you're trying to search for a value to find its position, iloc doesn't support this directly—you need to use value-matching operations first, like the ones above.

内容的提问来源于stack exchange，提问作者arun