如何将成对共发表相关性CSV表转换为Neo4j/Cytoscape可导入的节点边表

Got it, let's break this down step by step—turning your 165x165 co-publication correlation CSV into node and edge tables for Neo4j or Cytoscape is totally doable, whether you prefer coding or no-code tools. First, let's confirm we're on the same page about your CSV structure: I'm assuming it's a square matrix where the first row and first column are author names, and each cross-cell is the correlation coefficient between two authors.

Both Neo4j and Cytoscape rely on two core files:

• Node Table: A list of unique authors (each author is a node)
• Edge Table: A list of pairwise relationships, with the correlation as an attribute

Method 1: Python (Pandas) – Fast & Flexible #

This is the best approach for large tables like your 165x165 one, since it automates cleanup.

Generate Node Table #

import pandas as pd

# Load your correlation matrix CSV
df = pd.read_csv("your_correlation_matrix.csv", index_col=0)

# Extract unique author names for nodes
nodes = pd.DataFrame({
    "id": df.index.unique(),  # Unique ID for each node
    "label": df.index.unique()  # Display name (can match ID)
})

# Save node table
nodes.to_csv("authors_nodes.csv", index=False)

Pro tip: If you have extra author metadata (like affiliation), you can add those as additional columns here.

Generate Edge Table #

We need to flatten the matrix into a list of source → target pairs, and clean up duplicates/self-correlations:

# Convert wide matrix to long format
edges = df.stack().reset_index()
edges.columns = ["source", "target", "correlation"]

# Remove self-correlations (author paired with themselves)
edges = edges[edges["source"] != edges["target"]]

# Remove duplicate pairs (e.g., Alice→Bob and Bob→Alice) – keep only one direction
edges = edges[edges["source"] < edges["target"]]

# Save edge table
edges.to_csv("co_author_edges.csv", index=False)

This gives you a clean list of unique co-publication relationships with their correlation values.

Method 2: Excel (No-Code) #

If coding isn't your thing, Excel's Power Query can handle this:

Node Table #

1. Copy the first column of author names to a new worksheet
2. Go to Data → Remove Duplicates to get a unique list of authors
3. Add headers id and label (both can use the author name)
4. Save as authors_nodes.csv

Edge Table #

1. Select your entire correlation matrix (including headers and first column)
2. Go to Data → From Table/Range to open Power Query
3. Select the author name column, then go to Transform → Unpivot Columns → Unpivot Other Columns
4. Rename the columns: Attribute → target, Value → correlation, and the first column → source
5. Filter out rows where source = target (self-correlations)
6. To remove duplicate pairs, add a helper column with =IF([@source]<[@target],1,0), then filter for rows with 1
7. Export the cleaned table as co_author_edges.csv

Once you have your node and edge CSVs:

1. Move both files to Neo4j's import folder (found in your database directory)
2. Run these Cypher queries in the Neo4j Browser:

// Import author nodes
LOAD CSV WITH HEADERS FROM "file:///authors_nodes.csv" AS row
CREATE (a:Author {id: row.id, name: row.label});

// Import co-publication edges with correlation attribute
LOAD CSV WITH HEADERS FROM "file:///co_author_edges.csv" AS row
MATCH (a:Author {id: row.source}), (b:Author {id: row.target})
CREATE (a)-[:CO_PUBLISHED_WITH {correlation: toFloat(row.correlation)}]->(b);

Note: toFloat() converts the correlation value from a string to a numeric type for calculations later.

Cytoscape makes this super straightforward:

1. Open Cytoscape, then drag-and-drop both authors_nodes.csv and co_author_edges.csv into the window
2. Cytoscape will automatically detect:
   • id as the unique node identifier
   • source and target as the endpoints of each edge
   • correlation as an edge attribute
3. You can then adjust styles (e.g., make edges thicker for higher correlation values) via the Style panel on the right.

内容的提问来源于stack exchange，提问作者user3050520