igraph中调整聚类间距、命名聚类及优化网络绘图的技术问询

Hey there! Let's tackle each of your network visualization problems step by step with actionable code adjustments:

1. Increase Spacing Between Clusters #

The key here is to make clusters "stick together" more tightly, which naturally pushes other clusters further apart. Try these two approaches:

Option 1: Boost Intra-Cluster Edge Weights #

When adding edges for same-cluster nodes, crank up the weight to make the layout algorithm prioritize keeping cluster nodes clustered:

# Change weight from 500 to a higher value like 1000
Grouped.net = add_edges(Grouped.net, combn(GroupV, 2), attr=list(weight=1000))

Option 2: Tweak Layout Algorithm Parameters #

Adjust the layout_with_fr function to amplify repulsion between nodes outside the same cluster:

set.seed(567)
# Increase weight influence and node repulsion radius
LO = layout_with_fr(Grouped.net, 
                    weights = E(Grouped.net)$weight * 2,  # Make intra-cluster bonds stronger
                    repulserad = vcount(Grouped.net)^2.8)  # Push non-cluster nodes further away

2. Optimize Edge Display (Fix "Weird" Edge Appearance) #

With 1.2 million edges, overcrowding is the root cause. Clean up the edges with these tweaks:

Simplify Edge Visibility #

Reduce opacity, adjust width by weight, and remove curves for a cleaner look:

plot(net, layout=LO, edge.arrow.size=0, vertex.label=NA, asp=0, vertex.size=4,
     edge.width=ifelse(E(net)$weight > 1, 0.3, 0.1),  # Thicker edges for intra-cluster connections
     edge.color=adjustcolor("gray", alpha.f=0.15),  # Transparent edges to cut down on clutter
     edge.curved=FALSE)  # Straight edges are easier to parse

Filter Low-Weight Edges (Optional) #

If you don't need every edge, filter out low-weight ones to reduce noise:

# Keep only intra-cluster edges (weight >= 2)
filtered_net <- delete_edges(net, which(E(net)$weight < 2))
plot(filtered_net, layout=LO, ...)  # Use your existing plot parameters here

3. Add Cluster Names to the Plot #

Calculate the center of each cluster and overlay text labels using text():

# First plot the network as usual
plot(net, layout=LO, edge.arrow.size=0, vertex.label=NA, asp=0, vertex.size=4,
     edge.width=ifelse(E(net)$weight > 1, 0.3, 0.1),
     edge.color=adjustcolor("gray", alpha.f=0.15),
     edge.curved=FALSE)
legend(x=-1.5, y=-1.1, c("typeA","typeB", "typeC"), pch=21, col="#777777", pt.bg=colrs, pt.cex=2, cex=.8, bty="n", ncol=1)

# Calculate cluster center coordinates
cluster_centers <- aggregate(LO, by = list(Cluster = nodes$Clusters), FUN = mean)

# Add cluster labels to the plot
for (i in 1:nrow(cluster_centers)) {
  text(x = cluster_centers[i, 2], y = cluster_centers[i, 3],
       labels = paste("Cluster", cluster_centers[i, 1]),
       cex = 1.1, font = 2, col = "darkblue")
}

4. Move Rare typeC Nodes to the Top #

Manually shift the y-coordinates of typeC nodes upward to avoid them being buried:

# Find indices of typeC nodes (matches the "tomato" color in your colrs vector)
typeC_indices <- which(V(net)$type_num == which(colrs == "tomato"))

# Shift their y-coordinates upward (adjust the offset to fit your plot)
LO[typeC_indices, 2] <- LO[typeC_indices, 2] + 1.0

# Plot with the adjusted layout
plot(net, layout=LO, ...)  # Use your existing plot parameters

Alternative: Set Initial Layout Positions #

For more control, start typeC nodes at the top before running the layout algorithm:

# Create an initial layout matrix with random positions
initial_LO <- matrix(runif(vcount(Grouped.net)*2, -1, 1), ncol=2)
# Set typeC nodes to start at the top (y=1)
initial_LO[typeC_indices, 2] <- 1.0

# Use the initial layout in layout_with_fr
set.seed(567)
LO = layout_with_fr(Grouped.net, init = initial_LO, weights = E(Grouped.net)$weight * 2)

内容的提问来源于stack exchange，提问作者beginner