Lecture: Attributed Property Graphs

This lecture explores how to create, visualize, and filter graphs that store metadata within nodes and edges using Python's networkx library.

1. Creating a Property Graph

A property graph is a directed graph where nodes and edges can have key-value pairs (attributes). This is essential for modeling real-world entities like social networks or organizational structures.

import networkx as nx
import matplotlib.pyplot as plt

def create_and_visualize_properties_graph():
    # Initialize a Directed Graph
    G = nx.DiGraph()

    # Add nodes with 'role' and 'age' attributes
    G.add_node("Alice", role="Manager", age=30)
    G.add_node("Bob", role="Developer", age=25)
    G.add_node("Charlie", role="Designer", age=28)

    # Add edges with 'weight' attributes
    G.add_edge("Alice", "Bob", weight=5.0)
    G.add_edge("Alice", "Charlie", weight=2.0)
    G.add_edge("Bob", "Charlie", weight=3.5)

    return G

2. Visualization Techniques

To visualize attributes, we must map internal data to labels and layouts. The spring_layout is commonly used to position nodes based on a force-directed algorithm.

Node Labels: Combining names and roles.
Edge Labels: Displaying weights on the connections.
Layouts: Using pos = nx.spring_layout(G) to prevent overlap.

3. Filtering the Graph

Filtering allows us to isolate specific subsets of data. Below are the primary methods for querying graph attributes.

A. Filtering Nodes by Attribute

Using list comprehensions to find specific data points (e.g., nodes where age > 25):

senior_nodes = [node for node, attrs in G.nodes(data=True) if attrs.get('age', 0) > 25]

B. Creating Subgraphs

A subgraph is a new graph object containing a subset of the original nodes and their existing connections.

Important: G.subgraph() provides a view of the original graph, making it memory efficient.

Method	Logic	Common Use Case
Node Filtering	`G.nodes(data=True)`	Finding entities by role, age, or status.
Edge Filtering	`G.edges(data=True)`	Finding "strong" or "weak" connections via weights.
Subgraphs	`G.subgraph(list)`	Visualizing a specific department or team.

Summary of Implementation

When working with property graphs, always ensure your attributes are consistent across nodes to avoid KeyError issues during filtering.