Multi-line Time Series Chart in ggplot2

Question

I have a dataframe comprising two columns, 'host', and 'date'; which describes a series of cyber attacks against a number of different servers on specific dates over a seven month period.

Here's what the data looks like,

> china_atks %>% head(100)
                host       date
1     groucho-oregon 2013-03-03
2     groucho-oregon 2013-03-03
...
46 groucho-singapore 2013-03-03
48 groucho-singapore 2013-03-04
...

Where 'groucho-oregon', 'groucho-signapore', etc., is the hostname of the server targeted by an attack.

There are around 190,000 records, spanning 03/03/2013 to 08/09/2013, e.g.

> unique(china_atks$date)
  [1] "2013-03-03" "2013-03-04" "2013-03-05" "2013-03-06" "2013-03-07" 
"2013-03-08" "2013-03-09"
  [8] "2013-03-10" "2013-03-11" "2013-03-12" "2013-03-13" "2013-03-14" 
"2013-03-15" "2013-03-16"
 [15] "2013-03-17" "2013-03-18" "2013-03-19" "2013-03-20" "2013-03-21" 
"2013-03-22" "2013-03-23"
...

I'd like to create a multi-line time series chart that visualises how many attacks each individual server received each day over the range of dates, but I can't figure out how to pass the data to ggplot to achieve this. There are nine unique hostnames, and so the chart would show nine lines.

Thanks!

You dont have the number of attacks in the data, if yes where is it? — Rana Usman, Apr 01 '18 at 19:43
Each row appears to be a separate observation, so the number of attacks could be calculated by summarising this. — Michael Harper, Apr 01 '18 at 19:47
No the number of attacks will have to be counted in the actual data frame itself -- should this be my first step? — Edward Gargan, Apr 01 '18 at 19:48
Add more rows for another date, 2 rows won't cut, however my answer should work. — Rana Usman, Apr 01 '18 at 20:15

Rana Usman · Accepted Answer · 2018-04-01T20:05:32.357

3

Here's one way to do this.

First Summarize the count frequency by date.

library(plyr)
df <- plyr::count(da,c("host", "date"))

Then Do the plotting.

ggplot(data=df, aes(x=date, y=freq, group=1)) + 
  geom_line(aes(color = host))

Data

 da <- structure(list(host = structure(1:4, .Label = c("groucho-eu", 
    "groucho-oregon", "groucho-singapore", "groucho-tokyo"), class = "factor"), 
        date = structure(c(1L, 1L, 1L, 1L), .Label = "2013-03-03", class = "factor"), 
        freq = c(1L, 4L, 2L, 1L)), .Names = c("host", "date", "freq"
    ), row.names = c(NA, -4L), class = "data.frame")

edited Apr 01 '18 at 20:05

answered Apr 01 '18 at 19:51

Rana Usman

1,031
7
21

This is perfect, thanks! I had to change 'group=1' to 'group=host' for a smooth line between the points, otherwise it comes out as a sort of bar chart. – Edward Gargan Apr 01 '18 at 20:19
Good one. You can use `dplyr` which is next iteration of `plyr`. – MKR Apr 01 '18 at 20:21
1

@MKR Thanks! Could have done `dc <- da %>% group_by(host,date) %>% dplyr::summarise(freq = n())` but I think it would have been a bit longer. – Rana Usman Apr 01 '18 at 20:32
1

Absolutely correct!! But i meant you could have used `dplyr::count` – MKR Apr 01 '18 at 20:33
@MKR makes sense :) – Rana Usman Apr 01 '18 at 20:36

score 3 · Answer 2 · answered Apr 01 '18 at 20:54

ggplot2 library is capable of performing statistics. Hence, an option could be to let ggplot handle count/frequency. This should draw multiple lines (one for each group)

ggplot(df, aes(x=Date, colour = host, group = host)) + geom_line(stat = "count")

Note: Make sure host is converted to factor to have discrete color for lines.

Multi-line Time Series Chart in ggplot2

2 Answers2