如何更简便地通过统计单元格值计算Pandas共现矩阵？

Hey there! Your loop-based approach gets the job done, but we can leverage pandas' powerful vectorized operations to get the exact co-occurrence matrix you want in one concise line—no loops required. Let's walk through it step by step.

First, recap your data and goal #

Your original 0-1 DataFrame:

import pandas as pd
df = pd.DataFrame({'a' : [1,1,0,0], 'b': [0,1,1,0], 'c': [0,0,1,1]})

You need a matrix where each entry [i,j] counts how many times column j has a 1 when column i has a 1.

The One-Line Vectorized Solution #

This is exactly what a matrix dot product of the DataFrame with its transpose achieves:

co_occurrence_matrix = df.T @ df

Alternatively, using the explicit dot method (same result):

co_occurrence_matrix = df.transpose().dot(df)

Verify the Result #

Running this code gives you the exact matrix you wanted:

a  b  c
a  2  1  0
b  1  2  1
c  0  1  2

Why This Works #

Let's break down the logic:

• When you compute df.T @ df, the value at position (i,j) is the dot product of column i and column j from the original DataFrame.
• For your 0-1 data, the dot product counts the number of rows where both column i and column j are 1. This perfectly matches your rule: it sums up all the 1s in column j for every row where column i is 1.

This method isn't just shorter—it's also far faster for larger datasets, since pandas uses optimized vectorized operations instead of slow Python-level loops.

How It Compares to Your Original Code #

Your loop approach is functional, but this vectorized version is more readable, easier to maintain, and efficient. No need to juggle iteritems, groupby, or lambda functions here!

内容的提问来源于stack exchange，提问作者Edward