Research Projects
Smart-Miner: A
New Framework for Mining Large Scale Web Usage Data
In this project, we
are proposing a novel framework for Web Analysis Systems in order to improve commercial
web sites both structurally and contextually. Our framework provides feedback
and detailed analysis of commercial web site with several components from web usage mining system to decision
support system. The research
contributions and novelty of our project is listed below:
o Our framework has capability of generating its own enriched usage data containing semantic information and cookie information at client side. The cookie information included in web usage data contains all information about user browser interactions. By this way, we have all of the information about user action history in client side. In addition, we will exact referrer information from client side which enables us to improve the session reconstruction phase of web usage mining process in which we use link information between web pages. One of the novel works will be studied in this project is that, we will propose a new session reconstruction heuristics which extract maximal navigation paths in sessions by using referrer information in cookies. Unlike the several impractical session reconstruction methods, we do not need to extract site topology with crawler since we have all knowledge about web user at client side.
o Another contribution of the project is that we investigate
semantic navigation paths of the web user. Since we have semantically enriched
user data, we will explore frequent semantic paths which correspond to sequence
of terms in domain ontology. With this information, we will study on improving
web site contextually by using semantically enriched usage patterns.
o Another novel work that is included in our project is that
we will study inner page usage beside intra page usage. Unlike the ordinary web
usage mining methods focuses on navigation data between page web pages, we will
study on usage mining over component of single web page. We are going to track
user actions on predetermined parts of single web page for improving the
internal structure of the web pages.
o Decision Support system is the most important motivations of
this project. As it will be mentioned in the related work sections, there is
not enough study on these types of systems in web usage mining area. Also, the
current web analytic systems can not escape form being simple software that performs statistical
analysis on web site usage. With in the context of
this project, we will study on decision support system which will produce
suggestions about improvement of web site both structurally and semantically.
Decision support system will receive contextual information about navigation
patterns by using semantically enriched navigation paths. It also process these
semantically enriched and ordinary navigation patterns with decision rule sets
determined by web site owners. By this way, this system has capable of
producing suggestions about improvement of web site.
o We will also study on novel test system which enables to
track and chance predetermined parts of web site like web template for
analyzing the trend in the web site usage. With the help of this system, the
site owner can analyze the effect of specific component of web site over web
site usage.
o Another novel work is that we investigate the scalable
architecture for mining millions of usage data at server side. We are going to
investigate efficient implementation of pattern mining algorithms over map
reduce paradigm which is used for implementing complex algorithms over huge
amount of data, especially in search engine systems.
The
Architecture of the Framework is Given Below: