Information Technology Reference
In-Depth Information
this. The heuristic is based on the time between the appearance of a URL, and
the request for the resource it points to. This time is called think time, and it is
defined differently for browsers and users.
User Think Time
(
∆
e
): Length of time between a page download and a
hyperlink click (explicit request). User think time includes the browser's
parsing and rendering time.
Browser Think Time
(
∆
i
): Length of time between a page download
and an ancillary, automatic request (implicit request). Browser think
time includes browser parsing time.
Think time corresponds exactly to the length of time between matching URLs
in distinct connections. It can be represented by a label on each edge in a TCP
connection graph. An example of this is shown in Figure 4.
Fig. 4.
Think Times for Two Links between Two Connections
The figure shows two potential URL matches linking connections 1 and 2.
The first URL match implies an implicit request (an ancillary request made
automatically by a browser fetching embedded content), while the second implies
an explicit request (a request resulting from a human user click). Think times
are calculated for every URL match, including matches implying link-traversals
that never occurred. The marking of likely adjacencies is based on the length of
these think times.
The time oriented heuristic is a simple set of think time limits outside which
link traversals are deemed unlikely and removed. Link traversals (represented by
URL matches) are removed according to the following rules:
Implicit URL Match
(Browser Request): if think time
t
t
>∆
i
, remove.
Explicit URL Match
(User Request):
if think time
t
t
>∆
e
, remove.
Borrowing from the traditional sessionization techniques of web analytics [5], the
values of
∆
i
and
∆
e
are 20 seconds and four minutes respectively.
The heuristic is only applied to those connection nodes having an indegree
greater than one. That is, nodes with multiple incoming edges that imply the
node was linked-to from more than one other connection. An example is shown
in Figure 5.
