WEBVTT - This file was automatically generated by VIMEO

0
00:00:08.340 --> 00:00:12.720
Hi. In this video we will learn how to do multivariate analysis.

1
00:00:19.110 --> 00:00:23.210
We can plot the same scatter plot of count and service count variable, uh,

2
00:00:23.230 --> 00:00:27.330
but we can add a category variable, uh, to color the points

3
00:00:28.840 --> 00:00:32.690
like here. And now we see, um, other patterns.

4
00:00:35.120 --> 00:00:39.890
This area here is the Smurf attack. It has the, uh,

5
00:00:39.890 --> 00:00:42.930
high values of count and also of service count variable.

6
00:00:45.560 --> 00:00:48.610
This area here is the Saturn attack, uh,

7
00:00:48.610 --> 00:00:52.650
which has high values of count variable, uh, but low values, um,

8
00:00:53.350 --> 00:00:55.330
around zero of service count variable.

9
00:00:58.640 --> 00:01:01.340
The normal connections are centered around here

10
00:01:05.320 --> 00:01:08.140
and here. This part, uh, this should be the Neptune attack.

11
00:01:10.430 --> 00:01:14.370
Uh, noticed also that, uh, I use the parameter, uh,

12
00:01:14.550 --> 00:01:18.370
to plot our points a little bit smaller.

13
00:01:24.180 --> 00:01:28.360
We can do the same with a numeric variable, uh, which can also be, uh,

14
00:01:28.390 --> 00:01:32.290
used as a u or a color variable. Uh,

15
00:01:32.290 --> 00:01:34.170
here I wanted to see, uh,

16
00:01:34.230 --> 00:01:38.050
if there are any interesting patterns between destination host, service count,

17
00:01:38.290 --> 00:01:41.970
variable count around source bytes variable.

18
00:01:47.180 --> 00:01:48.640
We see again that uh,

19
00:01:48.750 --> 00:01:52.840
most of the values of source bytes variable are around zero.

20
00:01:53.810 --> 00:01:58.350
Uh, but here around 150 we see a little bit different color.

21
00:01:58.930 --> 00:02:01.070
And, uh, also here in the corner

22
00:02:12.810 --> 00:02:13.643
in scatterplot,

23
00:02:13.810 --> 00:02:18.430
we can add another variable which maps its values to size of the points.

24
00:02:18.660 --> 00:02:23.070
Like here in this example, uh, we see that, uh,

25
00:02:23.330 --> 00:02:28.230
the most of the connections are via U D P protocol and have

26
00:02:28.230 --> 00:02:32.190
count and service count variable around zero. But to have a better view,

27
00:02:32.530 --> 00:02:36.470
we should investigate closer because it might be just few outliers.

28
00:02:41.120 --> 00:02:45.170
When doing analysis of two categorical variables and one numeric,

29
00:02:45.350 --> 00:02:48.250
we can do a a faceted group box plot.

30
00:02:48.440 --> 00:02:50.850
Like here in this example here,

31
00:02:51.000 --> 00:02:55.930
each subplot is an attack and is distribution of count variable per

32
00:02:56.000 --> 00:02:56.833
flag

33
00:03:04.030 --> 00:03:07.040
type of charts for analysis of two ric or variables.

34
00:03:07.040 --> 00:03:09.240
And one numeric might be a heat map,

35
00:03:09.580 --> 00:03:13.480
but here you lose some information about the whole distribution. Uh,

36
00:03:13.480 --> 00:03:18.240
but when some, uh, summary statistic, uh, like mean or median is okay for you,

37
00:03:18.500 --> 00:03:19.360
uh, then it's fine.

38
00:03:22.810 --> 00:03:23.820
Here in the next example,

39
00:03:24.260 --> 00:03:28.780
I have plotted a density plot of count variable colored by flag variable

40
00:03:28.780 --> 00:03:32.080
categories and faceted by attack.

41
00:03:33.080 --> 00:03:35.120
Remember that before using pandas plot method,

42
00:03:35.260 --> 00:03:39.000
we had to calculate the probabilities manually, but the seaborn,

43
00:03:39.150 --> 00:03:41.160
this plot has attribute multiple.

44
00:03:41.860 --> 00:03:46.730
And here we can set the value field and the result

45
00:03:46.730 --> 00:03:47.563
will be the same.

46
00:03:52.910 --> 00:03:55.930
In our next video, we will learn, uh,

47
00:03:56.000 --> 00:03:59.170
some information about few other interesting charts.
