Naviz : User Behavior Visualization of Dynamic Page

1
Naviz User Behavior Visualization of Dynamic
Page

• Bowo Prasetyo, Iko Pramudiono et. al.
• Pacific-Asia Conference on Knowledge Discovery
  and Data Mining2002, Taipei, Taiwan
• Presented by Byung Hee Lee

2
Contents

• Introduction
• Motivation
• Related Works
• Naviz Navigational Behavior Visualization
• Known problems in previous works
• Terminology
• System overview
• Mining Algorithm
• Features
• Strategies
• Visualization of NTT i-Townpage served on i-Mode
• Consideration
• Conclusion

3
Introduction

• Webmaster must know how to improve structure of
  the hyperlinks
• By analysis of the visitor navigational behavior
  on the website
• By visualization of the discovered patterns
• It is useless that the whole pages are drawn on a
  screen
• Need a efficient visualization mecahnism
• Introduce a Java Applet based navigational
  behavior visualization tool, Naviz
• Consider the relationship between web attack
  visualization and this tool

4
Motivation
Session Identification
Data Cleaning
Web Usage Mining
Pattern Analysis
Web Server Log
Topology Scanning
Visualization
Anomaly Detection
Web Server URL
Site Topology
Profile-based IDS

• We need a whole web topology diagram for web
  attack pattern analysis

5
Related Works(1/7)

• WebViz
• Pitkow et.al. in 1994
• A tool for web log analysis and provides
  graphical view of web site
• Filtering domain name, directory name, start
  stop time, etc.
• Drawback
• Visualize the statistical property

6
Related Works(2/7)
7
Related Works(3/7)

• WUM Web Utilization Miner
• Spiliopoulou et.al. in 1998
• Human expert can dynamically specify the
  interestingness criteria for navigation pattern
• Drawback
• Lacks aesthetic visualization of the data

8
Related Works(4/7)
9
Related Works(5/7)

• VISVIP
• Cugini et.al. In 1999
• Visualize the paths taken through the site
• 3D display
• Drawback
• It needs client customization

10
Related Works(6/7)

• WebQuilt
• Hong et.al. in 2001
• It uses a proxy to log the activity

Client Browser
Proxy
Web Server
WebQuilt Log
11
Related Works(7/7)
12
Know Problems in Previous Works
Naviz Navigational Behavior Visualization

• Lack of capability to handle sequential patterns
• Lack of aesthetic visualization
• Do not consider about modern dynamic page
• Do not consider appropriate web traversal
  properties
• Hop count, visited page, time, place

13
Terminology
Naviz Navigational Behavior Visualization

• Dynamic page
• CGI, ASP,
• Page Class
• An abstraction of web page
• Instance
• A member of page class

Ex) Class - Time
Instance Day Night
14
System Overview(1/3)
Naviz Navigational Behavior Visualization

• Three-tier client-server application

15
System Overview(2/3)
Naviz Navigational Behavior Visualization

• Naviz Applet
• Display the visualization to the user
• Set the support value
• Set the confidence degree
• Naviz Servlets
• Interface between Naviz Applet and Server side
  programs

16
System Overview(3/3)
Naviz Navigational Behavior Visualization

• Pattern Repository
• Contain the discovered patterns
• Log Miner
• Discover traversals/path from web log data
• Association rule mining
• Sequential pattern mining
• Clustering
• Graphviz
• Draw the graph layout

17
Mining Algorithm
Naviz Navigational Behavior Visualization

• Association rule mining
• Session Support(A-gtB) session that contain
  A-gtB / total of session
• User Support(A-gtB) visitor that contain A-gtB
  / total of visitor
• Sequence pattern mining
• Confidence o a traversal path
• The probability of a visitor to visit the last
  page in the sequence of traversal path
• Ex
• Confidence(A-gtB-gtC-gtD) of contain A-gtB-gtC-gtD
  / of contain A-gtB-gtC

18
Features
Naviz Navigational Behavior Visualization

• Two operation mode
• Traversal diagram mode
• Traversal path mode
• Thickness of edge
• Represents support value
• Color of edge (range from blue to red)
• Represents confidence degree (low to high)

19
Strategies
Naviz Navigational Behavior Visualization

• Consideration of appropriate web traversal
  properties
• Hierarchization
• Grouping
• Class instance view
• Navigational behavior comparison
• Traversal path filtering
• Interactive environment

20
Visualization on NTT i-Townpage
21
Traversal Diagram
22
Visitor Success Path
23
Day/Night Success Path
24
Consideration
25
Conclusion

• Web log visualization using data mining
• Consideration of dynamic page and appropriate web
  traversal properties
• Need page clustering