Data Mining

成績

jdwang@asia.edu.tw, Room:I517, ext:1847

Time: Wednesday, 9:10am~12:00am, Room: A206

• Text Book

  • Data Mining: Practical Machine Learning Tools and Techniques (2nd Edition),

    (東華代理)

    Ian H. Witten & Eibe Frank. Morgan Kaufmann, June 2005,  ISBN 0-12-088407-0

    http://www.cs.waikato.ac.nz/ml/weka/book.html

    The Weka machine learning workbench.

  • ppt

• Reference Book

  • Data Mining: Concepts and Techniques (2nd Edition)

    Han& kamber, 2007, Morgan Kaufmann, ISBN 978-0-12-373905-6(東華代理)

  • Web Data Mining
    Exploring Hyperlinks, Contents and Usage Data
    Bing Liu, Springer, December, 2006

    • Data mining resource site: KDnuggets Directory

    • MDR: Mining Data Records from Web Pages

    • (CBA - Asociation rules based classification)

    • http://www.cs.uic.edu/~liub/WebMiningBook.html

• 資料探勘的發展與挑戰(作者：翁慈宗　成功大學資訊管理研究所)

  　

  ---

• 分組(2009.10.1前繳交分組名單,e-mail:施耀竣 aoc5202000@yahoo.com.tw )
  • 大學部2~4人一組，研究所2人一組
  • 上台報告、心得報告、貢獻度。
• Outline
• Pre-Request
  • Install Weka and Documentation.
  • Install LIBSVM and Practice.
    • SVM
      • A practical guide to support vector classification
      • A Practical Guide to Support Vector Classification_jdwang_2008_5_30.ppt
      • SVM Demo (by 廖益緯)
    • Download(Local in AsisUniversity )
      • LIBSVM(libsvm-2.88.zip)
      • LinearSVM
      • python(python-3.0.msi)
      • gunplot(gnuplot-4.2.2.tar.gz)
  • 笨蛋也可以用的libsvm
• Source Examples
  • Weka Datasets
  • UCI
  • UCI Knowledge Discovery in Databases Archive
  • Classification
  • 財團法人國家衛生研究院「全民健康保險研究資料庫
• Grading (最多遲交一週，逾期不收, 90%/per day)
  • 課堂表現(10%)：出席、提問、回答問題
  • Project 1(15%)(2009.10.16) :Data Preprocessing
  • Project 2(25%)(2009.11.6) :Association Rule & Decision Tree
  • Project 3(25%)(2009.12.4) : SVM classifier Rank
    • DataSet
      • http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/
    • LIBSVMDataSet_97_2_DataMining_jdwang.xls
      • 全部跑完(50%)(缺一組(-2))
        • (正確率最高:第一(+5),第二(+3),第三(+2),第四(+1))
      • 心得(30%);
  • Project 4(25%)(2010.1.8) :

    • Install Perl
      • Example. 2009_12_8_jdwangForSVMDataProcess
      • #Usage: perl ConvertToSVMDataFormat.pl <SrcFile> <NewFile> <Separator>
        perl ConvertToSVMDataFormat ClassLabelData.txt ClassLabelData.SVM ","
        　
      • #Usage: ComputeResultStatistics.pl <TestFile> <PredictFile>
        perl ComputeResultStatistics.pl iris.scale iris.scale.predict
      • Confusion Matrix

    • malibu: malibu is an open source, portable machine learning workbench written in C++. This collection of learning algorithms focuses on supervised learning problems. It includes both third-party and native implementations covering a number of classification algorithms and wrapper methods. malibu also encompasses the most complete set of validation algorithms, metrics, tests and graphs.
    • (2010.1.15,4:00pm前) 心得
      • Data Preprocessing (Categorical <=>Numerical)
      • How can you avoid the over-fitting problem?
        • Training Set, Validation Set, Testing Set, k-Fold cross validation
      • How can you  improve the classification accuracy?
        • Attribute Normalization, Classifier Combination
          • Accuracy,Precision,Recall,F1,...
      • Classifier Comparison ( DT, SVM, Linear SVM,  kNN, Neural Network, ....)
      • Data Transformation (PCA, LDA,...)
      • Cluster