SMDT

Appearance

Network Construction ​

SMDT provides powerful tools to construct various types of networks from your social media data. These functions generate networks over a specified period.

Supported network types include:

  • User Interaction: Direct interactions between users (e.g., Retweets, Mentions).
  • Entity Co-occurrence: Entities (e.g., Hashtags) that appear together in the same post.
  • Bipartite Networks: Connections between two different node types (e.g., Accounts and Hashtags).

Prerequisites ​

Ensure you have a database connection ready.

python
from smdt.store.standard_db import StandardDB
from smdt import networks
from datetime import datetime
import os

# Connect to your database
# initialize=False is recommended when just reading data
db_name = os.getenv("DEFAULT_DB_NAME", "smdt_db")
db = StandardDB(db_name, initialize=False)

Entity Co-occurrence Networks ​

This type of network connects entities that appear together in the same post. A common use case is a Hashtag Co-occurrence Network, where edges represent hashtags used in the same context.

python
# Generate a Hashtag Co-occurrence Network
# for a specific 24-hour window
ht_net = networks.entity_cooccurrence(
    db,
    entity_type="HASHTAG",
    start_time=datetime(2023, 5, 14),
    end_time=datetime(2023, 5, 15),
    min_weight=5,  # Only include edges with at least 5 co-occurrences
)

print(ht_net.meta)
print(ht_net.edges)

Parameters ​

  • entity_type: The type of entity to analyze (e.g., 'HASHTAG', 'USER_TAG', 'LINK').
  • start_time / end_time: Filter data to this time range.
  • min_weight: Filter out edges with fewer than N co-occurrences to reduce noise.

Bipartite Networks ​

Bipartite networks connect two different types of nodes. For example, an Account-Hashtag Network connects users to the hashtags they have used.

python
# Build account–hashtag bipartite network
user_ht = networks.bipartite(
    db,
    left="account",
    right="hashtag",
    start_time=datetime(2023, 5, 14),
    end_time=datetime(2023, 5, 15),
    weighting="count",
)

print(user_ht.meta)
print(user_ht.edges)

Parameters ​

  • left: Must be 'account' or 'post'.
  • right: Any valid entity type (e.g., 'HASHTAG', 'LINK').
  • weighting: 'count' (number of connections) or 'binary' (1 if connected).

User Interaction Networks ​

These networks represent direct social interactions. For example, a Retweet Network connects users who retweeted others.

python
# Build a Retweet Network (Who retweeted whom?)
rt_net = networks.user_interaction(
    db,
    interaction="QUOTE", # or "SHARE", "COMMENT", "ReTweet" depends on platform mapping
    start_time=datetime(2023, 5, 14),
    end_time=datetime(2023, 5, 15),
    min_weight=2
)

print(f"Interaction edges: {len(rt_net.edges)}")

Working with Results ​

All network functions return a NetworkResult object containing pandas DataFrames.

python
# The edges are a standard DataFrame
df = ht_net.edges
# Columns: ['source', 'target', 'weight', ...]

# You can easily export to CSV
df.to_csv("hashtag_network.csv", index=False)

# Or convert to NetworkX for analysis
import networkx as nx
G = networks.to_networkx(ht_net)
print(nx.info(G))