Introduction
• A well-designed website is beneficial to visitors.
• Navigation and search
• Structure of hyperlinks
• Definition of website navigability
• Aside from perceptual measurements, a data-driven approach is also
can be utilized to evaluate navigability of websites
• Limitations of navigability processed by other scholars in the past
• The objectives of the paper
• Three metrics : power, efficiency, directness

Literature Review
• Website navigation and navigability
 Critical influence of navigation
 Navigation systems, important means
 Nuance between navigation and navigability
• Measuring navigability with web data
 Broad classification
 Web content mining
 Web structure mining
 Web usage mining

Theoretical Foundations
• Information foraging theory
 It extends the optimal foraging theory
 Very likely to modify browsing strategies
• Information-processing theory
 People process information via many aspects
• Visitors make judgments about their traversing paths
• What they care doesn’t merely contain the likelihood of locating target
information.

Method and Metrics for
Measuring Navigability
A Web Mining–Based
Method for Measuring
Navigability
Steps
1. Web log preprocessing：Cleaning, session identification, session
completion.
2. Web site parsing：Parsing focal sites
3. Web page classification：Content pages and index pages
4. Access pattern mining：Frequently accessed sequences of content
pages as proxies for information-seeking targets
5. Hyperlink Structure representation：A distance matrix

Data-Driven Metrics for
𝑝 𝑘 =
𝛽
2𝜋𝑘3 exp
−𝛽 𝑘−𝛼 2
2𝛼2 𝑘
𝑘 = 1,2, … ,
𝐺 𝐼 = ∀𝑘≥𝐼 𝑝(𝑘) 𝐼 = 1,2, …
𝐺 𝐼 =
1 if 𝐼 = 1
G 𝐼 − 1 − 𝑝 𝐼 − 1 if𝐼 > 1

Power
• 𝑈 = 𝑢𝑖 , 𝑖 = 1,2, … , 𝑛
• 𝑢𝑖 =< 𝑝𝑖,1, 𝑝𝑖,2, … , 𝑝𝑖,𝑚 >, where 𝑝𝑖,𝑗 is the jth content page in 𝑢𝑖, 𝑗 =
1,2, … , 𝑚
• 𝑅 𝑢𝑖|𝑝𝑠 = 𝐺 𝑑 𝑝𝑠, 𝑝𝑖,1 𝑗=2
𝑚
𝐺 𝑑 𝑝𝑖,𝑗−1, 𝑝𝑖,𝑗 if 𝑝𝑠 ≠ 𝑝𝑖,1,
otherwise 𝑅 𝑢𝑖|𝑝𝑠 = 𝑗=2
𝑚
𝐺 𝑑 𝑝𝑖,𝑗−1, 𝑝𝑖,𝑗

Power
• 𝑅 𝑢𝑖 = ∀𝑝 𝑆
𝑃(start of seeking for 𝑢𝑖 = 𝑝𝑠)𝑅 𝑢𝑖|𝑝𝑠
• Introducing weight
 𝑤 𝑢𝑖 =
𝑣(𝑢 𝑖)
∀𝑢∈𝑈 𝑣(𝑢)
 𝑅 𝑈 = 𝑖=1
𝑛
𝑤 𝑢𝑖 𝑅(𝑢𝑖)

Efficiency
• 𝑄 𝑢𝑖|𝑝𝑠 =
𝑚𝛾−min(𝑑 𝑝 𝑠, 𝑝 𝑖,1 + 𝑗=2
𝑚
𝑑 𝑝 𝑖,𝑗−1, 𝑝 𝑖,𝑗 , 𝑚𝛾)
𝑚(𝛾−1)
• 𝑄 𝑢𝑖|𝑝𝑠 =
(𝑚−1)𝛾−min( 𝑗=2
𝑚
𝑑 𝑝 𝑖,𝑗−1, 𝑝 𝑖,𝑗 , (𝑚−1)𝛾)
𝑚(𝛾−1)
, if 𝑝𝑠 = 𝑝𝑖,1
• 𝑄 𝑢𝑖 = ∀𝑝 𝑠
𝑃(start of seeking for 𝑢𝑖 = 𝑝𝑠) 𝑄 𝑢𝑖|𝑝𝑠
• 𝑄 𝑈 = 𝑖=1
𝑛
𝑤 𝑢𝑖 𝑄(𝑢𝑖)

Directness
• 𝐿 𝑢𝑖|𝑝𝑠 =
𝑚𝛿−min(𝑁 𝑝 𝑠,𝑝 𝑖,1 + 𝑗=2
𝑚
𝑑 𝑝 𝑖,𝑗−1, 𝑝 𝑖,𝑗 , 𝑚𝛿)
𝑚(𝛿−1)
if 𝑝𝑠 ≠ 𝑝𝑖,1
• 𝐿 𝑢𝑖|𝑝𝑠 =
(𝑚−1)𝛿−min( 𝑗=2
𝑚
𝑑 𝑝 𝑖,𝑗−1, 𝑝 𝑖,𝑗 , (𝑚−1)𝛿)
(𝑚−1)(𝛿−1)
if 𝑝𝑠 = 𝑝𝑖,1
• 𝐿 𝑢𝑖 = ∀𝑝 𝑠
𝑃(start of seeking for 𝑢𝑖 = 𝑝𝑠) 𝐿 𝑢𝑖|𝑝𝑠
• 𝐿 𝑈 = 𝑖=1
𝑛
𝑤 𝑢𝑖 𝐿(𝑢𝑖)

Implementation and Illustrations
• An archetype system was established.
• SpidersRUs was used to parse a website.
• Two sites
 A
 3840 content pages
 437 index pages
• Web logs were gleaned over four weeks.
 A：35,966,494 records; 732,321 sessions
 B：32,170,062 records; 555,299 sessions
 B
 3738 content pages
 380 index pages

•
• The threshold was at first set at 0.05%, then its value was increased
with 0.025% in the range from 0.05% to 0.175%.
•

• The distances of power and efficiency on B is great on A.
• The directness distances between A and B are smaller than that of
power and efficiency.
• According to the proposed metrics, A has higher navigability than B
• The assessment of the proposed metrics and the prevalent metrics

Evaluation Study and Data Collection
Study design
• A group of people were recruited.
• The significance of users’ familiarity was addressed.
• Four experimental conditions were created
Tasks
• A pretest was conducted.
 Content pages are more likely to constitute information-seeking
targets.
 Key access sequences identified from Web logs are consistent with
users’ common information-seeking needs, desires, and interests.

Evaluation Study and Data Collection
Participants
• Business undergraduate students enrolled in similar information
systems or operations classes in both universities.
• Each participant received $10 for his or her time and efforts.
Measurements
• Three measures: task success rate, task time, and the number of clicks.
• Participants had up to 4 minutes to complete each task.
• Cognitive-processing load
Data collection
• A quite formal way

Data Analyses and Results
• A pilot study, 39 undergraduate students
• An evaluation study with 248 participants
• Comparison of user performance and assessments between A and B
• Comparison of user performance by separating tasks related to
complexity
• Performance of the participants from each university
• An ex post facto comparison
• Further examination of the proposed metrics

Extensions to Proposed Metrics
• A scale factor can be added while evaluating a larger website.
• The metrics can be extended with the combined use of search engine.
• Integration of three metrics as a holistic measure
3
1
𝑅(𝑈)
+
1
Q(𝑈)
+
1
𝐿(𝑈)
=
3𝑅 𝑈 Q 𝑈 𝐿 𝑈
Q 𝑈 𝐿 𝑈 + 𝑅 𝑈 𝐿 𝑈 + 𝑅 𝑈 Q 𝑈
= 𝑂(𝑈)

Discussion
• Three data-driven metrics and a viable method were presented.
• The method can be used continuously for supervising a website’s
navigability
• A method by Liu et al. is suggested for gleaning data (Web log).
• It helps improve hyperlink structure designs of websites
• Limitations
• Different structures of websites may not fit to the results
• Spiders and page parsers‘ utilities are limited.
• Test of different scenarios
• More factors can be introduced to perfect the method

Conclusion
• Three data-driven metrics were presented.
• By integrating appropriate Web mining techniques, a method
cooperated the metrics was created.
• The verification of the metrics and method.
• Users’ perception corresponds to navigability measured using the
methods established by the authors

Comment
• The article clearly and laconically expresses the idea and concept with
the existing theories.
• Vivid examples following many statements which we as post-graduate
students can look upon.
• A host of demonstrations below on many pages provide necessary
assistance for lay people
• I think navigability won’t be only one factor that may affect a website
access ratio.

A data driven approach to measure web site navigability

More Related Content

What's hot (20)

Similar to A data driven approach to measure web site navigability (20)

More from Shu-Jeng Hsieh (12)

Recently uploaded (20)

A data driven approach to measure web site navigability

Editor's Notes