Personalized Course Navigation Based on Grey Relational Analysis

1
Personalized Course Navigation Based on Grey
Relational Analysis

• Han-Ming Lee, Chi-Chun Huang, Tzu-Ting Kao
• (Dept. of Computer Science and Information
  Engineering, National Taiwan University of
  Science and technology)

Presented by Sharon HSIAO Feb.23.2007
2
agenda

• Introduction/motivation
• Course Recommending Procedure
• Results Evaluation
• Suggestions

3
Introduction

• Aim to provide a personalized information
  recommendation system that dynamically reflects
  users interests
• Focus model users interests without explicit
  rating
• Content-based personalized technique
• WGRA (Weighted Grey Relational Analysis)
• Coursebot System distance learning system

4
Coursebot

• Agent-based system
• Gather course materials from internet
• Make intelligent learning recommendations
• Classification methods style retrieval
  techniques to extract features

5
5 components wrapper agent, course constructor,
query agent, interface agent, scheduler
6
Coursebot 5 components

• Wrapper Agent collect course material webpages,
  then classify them by topics in given subjects
• Course Constructor organize webpages from course
  database as the materials in response to users
  queries
• Query Agent retrieve and expand the query from
  db
• Interface Agent learns profiles based on users
  browsing behavior
• Scheduler regularly command Wrapper agent to
  collect materials

7
Personalized Course Navigation

• Learning and ranking based on user profiles
• Use WGRA measure to analyze user preferences

8
How does it actually work?

• Interact (Query Agent, Interface agent)
• Time spent on a page (gt15 mins is discarded)
• Length of each page in bytes is recorded
• Feature vector is used (A Df1,f2,,fm)
• Course Display (Query Agent, Course constructor)
• Rank by revised user profiles and learning
  schedule of different topic (predefined)
• No ranking for 1st time user

9
WGRA (weighted Grey Relational Analysis)

• To analyze degrees of relevance among a visited
  page

Row individual feature of the document Column
the degree of Grey relation assigned to the
feature fi between each doc. in Ti and D1
10
The higher degree ?i1 between Di D1 means That
these two docs are related to each other
A longer visit to a given page, the user Probably
has higher interest
According to the interests of the doc(browsing
time length of page), apply adjustment to WGR
grade vector
11
Example
12
Experiment results

• 7 topics Neural Networks
• 1032 related webpages (spider)
• 128 features (style retrieval)
• 69 Ratings (graduate students who had taken NN)

13
(No Transcript)
14
conclusion

• The proposed method was not significantly
  different from other algorism
• User profiles are easily maintained
• Low complexity
• Ease to add knowledge
• suitable for online personalized analysis

15
Suggestions/notes

• Users are restricted to receiving documents
  similar to related items seen previously by other
  user
• Users interests concerning various course
  materials can be easily modeled