Exploring temporal structure

Case Studies: Network Analysis in R

Edmund Hart

Instructor

Are important products always important?

# Get unique Dates
d <- sort(unique(amzn_raw$date))

# Create graph from first date
amzn_g <- graph_from_data_frame(
  amzn_raw %>% 
  filter(date == d[1]) %>%  
  select(from, to), directed = TRUE
)

Are important products always important?

# Find products that are "important"
high_out_degree <- degree(amzn_g, mode = "out") > 2

low_in_degree <- degree(amzn_g, mode = "in") < 1

important_nodes <- high_out_degree & low_in_degree

imp_prod <- V(amzn_g)[importnant_nodes]


# Store as a data frame to later join on
tmp_df <- data.frame(imp_prod = as.numeric(names(imp_prod)))

## Create list to hold output
time_graph <- list()

## Create a 2x2 layout for plots and increase margins
par(mfrow = c(2, 2), mar = c(1.1, 1.1, 1.1, 1.1))

## Loop over the data to build 
for(i in 1:length(d)){

  ## Create a data frame at each time stamp
  ip_df <- amzn_raw %>% 
    filter(date == d[i]) %>% 
    right_join(tmp_df, by = c("from" = "imp_prod")) %>% 
    na.omit()

  ## Create an igraph object from that data frame
  time_graph[[i]] <- ip_df %>% 
    select(from, to) %>% 
    graph_from_data_frame(directed = TRUE)

  ## See what important vertices look like by date
  plot(time_graph[[i]], main = d[i])   }

Let's practice!

Case Studies: Network Analysis in R