Information Technology Reference
In-Depth Information
as soon as new cases are available. The tentative underlying conceptual structure of
the access pattern is visually presented to the user. We have developed two
approaches for clustering access patterns. Both are based on approximate graph
subsumption. The first approach is based on a divide-and-conquer strategy whereas
the second is based on a split-and-merge strategy which better allows to fit the
hierarchy to the actual structure of the application, but requires more complex
operations. The first approach uses a fixed threshold for the similarity values. The
second approach uses an evaluation function for the grouping of the cases.
5.2.5 Reporting Tools
Although the outcome of the data mining component is a set of rules or a description
of the clusters which can be directly used to control the functional components of the
website or directly incorporated into the user modeling component. We also prefer to
report the results of the data mining process in a form a system administrator or
marketing person can use for further review of the results. Therefore we will integrate
visualization components into our system that allow to visualize the resulting decision
tree, the hierarchical representation of the conceptual clusters, and the statistics for the
event marketing.
5.2.6 Knowledge Repository
In the knowledge repository are stored individual user profiles and product models
and preferences. The individual user profile is created by the user with the help of the
registration component of the user interface. It can be updated by the user itself or
electronically by the data mining component after having analyzed the user data when
visiting the website.
6 Conclusions
We have introduced a new architecture extending an e-shop into an intelligent e-
marketing and selling platform which can adapt to user needs and preferences. The
data which can be accessed during a user session as well as the method for analysing
these data play an important role for achieving this goal. Therefore, we have reviewed
the basic data that can be created during a customer session. Based on the kind of data
and the wanted output the data mining methods are selected. We have reviewed the
basic data mining methods and given an overview on what kind of method is eligible
for the considered result. We have identified two types of data mining methods useful
for our first set up of the intelligent e-shop. These are classifications based on
decision tree induction and conceptual clustering. With these methods we can solve
such problems as learning the user model, web usage mining for web site
organization, campaign management, and event monitoring. The data might be
labeled or might not have a label. In the latter case clustering is to use to learn similar
groups and label them. Recently, we have continued to develop and implement the
methods for decision tree induction and conceptual clustering. Each method will be
implemented as a component with standard input and output interfaces that allows to
assemble the components as far as will be needed for the particular e-shop.
