Loading...
Please wait, while we are loading the content...
Similar Documents
. Introduction 1.1 Problem Statement 1.2 Solution Approach 2. Related Research 2.1 Introduction 2.2 Implicit Systems 2.2.1 the Tapestry Text Filtering System
| Content Provider | Semantic Scholar |
|---|---|
| Abstract | Nowadays many research organizations are working on developing intelligent agents on the World Wide Web that can learn a user's interest and find information in the World Wide Web based on the user's profile. Goecks et al., 2001) Some researches found relationships between a user's behavior and user's interest level while others used a content analyzer to build predictive models to predict user's interest level on a web page when a user visits the web page. We think about building predictive models only using user's behavior with the user's explicit rating. There are some problems to predict a user's interest level to record and analyze the relationship between a user's interest level and the user's behavior on the World Wide Web. To monitor a user's interest indicators such as the number of the mouse clicks while a user uses a web browser, we need to have a monitoring software program to see what a user does on the web browser and how much he likes the web page. In addition to that, after we find the user's interest indicators general enough to build predictive models, we are able to build predictive models that can predict a user's interest level according to the user's behavior. We can think about building predictive models using regression analysis and neural networks. The two main motivations follow: • Can we record a user's interest indicators general enough to build predictive models? • Can we build predictive models to predict a user's interest level only using the user's behavior after the user leaves the web page? In this study, we show how we collect each user's interest indicators and build predictive models to predict a user's interest level by recording and analyzing only user's behavior with the user's explicit rating on the World Wide Web. Research has shown that a user's behavior is related to his interest level such as reading time spent on the web page. However, there were some technological problems and limitations in recording and analyzing a user's behavior general enough to build predictive models to predict how much a user is interested in the web page. Two main hypothesis of this thesis follows: • There are users' interest indicators that tell a user's interest level on a web page • We can build predictive models that predict a user's interest level using the user's behavior In addition to that, we hypothesized that … |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://cs.fit.edu/media/TechnicalReports/cs-2002-05b.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
