Database Reference
In-Depth Information
given web pages, along with statistics on how often each page of each route has been
accessed. The patterns and statistics provide rules with which the analyst can determine the
output of coincidence. By studying this route more closely and comparing it to other routes
crossing it, the web designer can detect pages that are not properly designed or linked and
redesign them.
Restructuring a Web Site According to the Mining Results
Path traversal patterns discovery helps the web designer in improving the design of web
sites. Detecting user navigation paths and analyzing them results in a better understanding
of how users visit a site, identifi es users with similar information needs, or even improves
the quality of information delivery in WWW using personalized web pages.
Also, the sequence of requests by visitors helps predict next requests or popular requests
for given days, and thus improves the network traffi c by caching those resources, or by
allowing the clustering of resources in a site based on user motivation.
Improving Customization
Customization involves learning about an individual user's preferences or interests based
on user access patterns. Thus, customization aids in providing users with pages, sites and
advertisements that are of interest to them. It may also be possible for web sites to automatically
optimize their design and organization based on observed user access patterns.
The Impact of Web Advertisements using OLAM
The openness is one of the WWW's biggest advantages. It introduces risk for information
security but is also a huge issue in users analysis; not because of its vast volume in eyeball
count but its random and extreme pattern of click and tick sequence on the company/
institution's web site. We have therefore embedded the value of Confi dence and Support
level to accommodate these issues of boundary-lessness in our OLAM approach. Although
the user of the prototype sets these two values solely based on heuristics, the criterion of
optimality in different business domains must always associate with their expertise knowledge.
As such, we restrict the user type of our prototype within the Sales Management team of
a company or the Public Relation team of an institution whose web site is undergoing the
mining process for increasing sales or promoting company images.
We believe that any e-customer can come to the web site and complete an e-service
process from beginning to end in a user-friendly and intuitively correct manner. We need to
encapsulate all our web site surfers' online experience to discover the knowledge of customer
behavior. OLAM creates a list of association rules for each targeted web page determined
by the user. The web pages tick sequences are represented in path traversal patterns. These
patterns are analyzed to discover users' preferences. The preferred web page(s) can be
identifi ed and categorized by Internet surfers and/or e-customers, as shown in Figure 14.
In Figure 14, web page #1 is the most frequently accessed web page. Web advertisement
then considers placing on this page. This is the most straightforward way without much need
of data mining technique. The OLAM approach offers more. Assume web page #8 is the
function page for registration as clients. All UIDs identifi ed in their tick sequence with the
visit to web page #8 are grouped as targeted e-customers. There may have been many routes
that could link to web page #8. Some users may have sent their registration and placed an
order/enquiry, whereas some may have skipped away. The unsuccessful cases are the target
