Lecture: Prof. Hahn-Ming Lee Student: Ching-Hao Mao [email protected]

Fuzzy Final HomeworkSystem ImplementationSelected paper: Fuzzy integration of structure adaptive SOMs for web content mining, Fuzzy Sets and Systems 148 (2004) 43–60

Lecture: Prof. Hahn-Ming LeeStudent: Ching-Hao Mao

[email protected]

Outline Introduction Proposed method in selected paper Implementation Conclusion References

Introduction In this report, we implement Kim and Cho’s paper

appear on Fuzzy Set and System in 2004 User profile represents different aspects of user’s

characteristics The author proposed an ensemble of classifiers

that estimate user’s preference using web content labeled by user as “like” or “dislike”

Introduction- Preview Studies [2]

Feature Selection Method Properties

Feature selection methods such as Information Gain, TFIDF, and ODDS ratio have different properties

TFIDF does not consider class values of documents when calculating the relevance of features while information gain uses class labels of documents

Odds ratio uses class labels of documents but they find useful features to classify only one specific class

Overview of the proposed method in [1]

Classification

TFIDF, Information Gain,ODDS Ratio

Structure Adaptive SOM

Training SASOM’s using different feature sets

Fuzzy Integral

Hot

Cold

or

Data Set Description UCI Syskill & Webert data (http://kdd.ics.uci.edu) Contain the HTML source of web pages plus the

ratings of a single user on these web pages The web pages are on four separate subjects

Bands- recording artists (Implement in this report) Goats (Implement in this report) Sheep BioMedical

Implementation Coding Java (J2SE 1.5) program for

preprocessing, feature selection (TFIDF and ODDS Ratio), and Fuzzy Integral mechanism

Using Weka for Feature Selection (Information Gain) and Classification

This report not successfully program SASOM…

Implementation-preprocessing

UCI Syskill & Webert data

ExtractHTMLContent.java

Pure Text without Anchor Text

Bands.txt

After Stopword and Porter Stemmer

Bands_Stopword.txtBands_Porter.txt

Implementation- Feature Selection In Bands, 61 dataset E.g. Attribute

Number: 5436->32

Information Gain TFIDF ODDS Ratio

0.1435 1411 mother0.1435 4109 writes0.1054 49 places0.1054 855 letter0.1054 3883 movement0.1054 1464 stories0.1054 3856 synthesizer0.1054 2568 songwriters0.0962 4643 singer0.0937 50 america

seaacidprogramminginnovativelettermethodmembersbleedconcentratedmother

osswildculturesvehementlysmokingdefinebookchargelibraryhand

Implementation- Fuzzy IntegralFuzzy measure of classifiers that are determined subjectively [1]

Bayes Classifier b1,b2,b3

b1=0, b2=1, b3=0 0.99

FuzzyIntegral.java

(g1,g2,g3)0.99,0.99,0.99) (0.01,0.01,0.99)(b1,b2,b3) Result (b1,b2,b3) Result

(0,1,0) 0.99 (0,0,1) 0.01

(1,1,1) 0.99 (0,1,1) 0.01

(0,0,0) 0,99 (0,0,0) 0.01

Conclusion Fuzzy integral provides the method of measuring

the importance of classifiers subjectively, especially in semi-supervised learning method

The method based on fuzzy integral can be effectively applied to web content mining for predicting user’s preference as user profile

Fuzzy Integral maybe can apply into my research area to integrate expert or user’s knowledge

References1. Kyung-Joong Kim, Sung-Bae Cho, Fuzzy integration of structure adaptive

SOMs for web content mining, Fuzzy Sets and Systems 148 (2004) 43–602. Pazzani M., Billsus, D., Learning and Revising User Profiles: The

identification of interesting web sites, Machine Learning 27 (1997), 313-331

3. http://kdd.ics.uci.edu/databases/SyskillWebert/SyskillWebert.data.html