require(igraph)
require(dplyr)

# Simulate some smaller scale data:
set.seed(123)

# fake IP addresses, assume counts follow normal dist (some heavy browsers, some light browsers)
IPcounts <- round(rnorm(100,mean=10,sd=3))

# Repeat IPs by fake counts
IP <- NULL
for(i in 1:length(IPcounts)){
  IP <- append(IP,rep(i,IPcounts[i]))
}

# fake domains, assume inverse popularity of visits (some very popular domains, many less popular domains)
domainCounts <- round(1000*(1/c(1:10000))+1)

# repeat domains by popularity
domain <- NULL
for(i in 1:length(domainCounts)){
  domain <- append(domain,rep(i,domainCounts[i]))
}


# randomize the domain list
domain <- sample(domain)
# randomly sample IPs to match the domain list's size
IP <- sample(IP,length(domain),replace=T)

# form the data frame
testData <- data.frame(domain,IP)

# transpose the incidence matrix to form a one-mode adjacency matrix
testData <- tcrossprod(table(testData))

# Use igraph to form up a graph of the data above
testNet <- graph.data.frame(testData, directed = F,)

# See nodes
V(testNet)

# See edges
E(testNet)

# inspect degrees
degree(testNet)

# Change node size to vary with degree
V(testNet)$size<-0.5

V(testNet)$color<-'white'

# Plot the result
par(mai=c(0,0,1,0))
par(bg = 'black')
set.seed(123)
plot(testNet,
     layout=layout.fruchterman.reingold,
     main='SmallWeb',
     vertex.frame.color='white',
     vertex.label = NA,
#      edge.curved=F,
     edge.arrow.size = 0.2, # for directed plot
     edge.color = hsv(0.5,0.9,1,alpha=0.3)
)


# Remove low degree connections
bad.vs<-V(testNet)[degree(testNet)<2] #identify those vertices part of less than 2 edges
testNet<-delete.vertices(testNet, bad.vs) #exclude them from the graph

set.seed(123)
plot(testNet,
     layout=layout.fruchterman.reingold,
     main='SmallWeb',
     vertex.frame.color='white',
     vertex.label = NA,
     #      edge.curved=F,
     edge.arrow.size = 0.2, # for directed plot
     edge.color = hsv(0.5,0.9,1,alpha=0.3)
)