Clean and trim network to create cohesive community

Iteratively refines a network by removing isolated users and applying follower thresholds to create a cohesive community where every user both follows and is followed by others within the network. This produces cleaner network visualizations and more meaningful community analysis.

Usage

trim_net(net, threshold)

Arguments

net

Tibble. Network edge list with follow relationships containing:

actor_handle

Character. User who is following another

follows_handle

Character. User being followed

threshold

Numeric. Inclusion threshold for users:

>= 1

Minimum number of followers within the network required

< 1

Minimum proportion of network that must follow the user

Value

Tibble. Cleaned network with the same structure as input but with:

• All duplicate edges removed

• Invalid handles and NA values excluded

• Users below follower threshold removed

• Only users who both follow and are followed within the network

The resulting network forms a more cohesive community for analysis.

Details

The trimming process works iteratively:

1. Remove duplicate connections and invalid/missing handles

2. Count followers for each user within the network

3. Remove users below the specified threshold

4. Keep only users who appear as both followers and followed

5. Repeat until the network stabilizes

This creates a "core community" where every member has meaningful connections within the group, improving visualization quality and analytical insights.

See also

expand_net, build_network, add_metrics

Other network-building: build_network(), expand_net(), init_net()

Examples

if (FALSE) { # \dontrun{
# Build an initial network
auth <- get_token("your.handle.bsky.social", "your-app-password")
initial_net <- init_net(
  c("scientist1.bsky.social", "researcher2.bsky.social"),
  c("science", "research"),
  auth$accessJwt
)

# Expand the network
expanded <- expand_net(
  net = initial_net,
  keywords = c("science", "research", "academic"),
  token = auth$accessJwt,
  refresh_tok = auth$refreshJwt,
  threshold = 5,
  max_iterations = 10
)

# Before trimming - show network size
cat("Before trimming:")
cat("  Unique actors:", length(unique(expanded$net$actor_handle)))
cat("  Unique follows:", length(unique(expanded$net$follows_handle)))
cat("  Total connections:", nrow(expanded$net))

# Trim to create cohesive community (need 10+ internal followers)
trimmed_net <- trim_net(expanded$net, threshold = 10)

# After trimming - show the refined network
cat("After trimming:")
cat("  Unique users:", length(unique(trimmed_net$actor_handle)))
cat("  Total connections:", nrow(trimmed_net))

# Verify all users both follow and are followed
actors <- unique(trimmed_net$actor_handle)
follows <- unique(trimmed_net$follows_handle)
cat("Network cohesion check:")
cat("  All actors also appear as follows:", all(actors %in% follows))
cat("  All follows also appear as actors:", all(follows %in% actors))

# Compare network density before and after
users_before <- length(unique(c(expanded$net$actor_handle, expanded$net$follows_handle)))
users_after <- length(unique(c(trimmed_net$actor_handle, trimmed_net$follows_handle)))
density_before <- nrow(expanded$net) / (users_before * (users_before - 1))
density_after <- nrow(trimmed_net) / (users_after * (users_after - 1))

cat("Network density improved from", round(density_before, 4), "to", round(density_after, 4))

# Use proportional threshold for smaller networks
# Require at least 20% of network to follow each user
proportional_trim <- trim_net(initial_net, threshold = 0.2)
} # }