Data Mining: the Practice Seminar

1
Data Mining the PracticeSeminar

• Lehrstuhl für maschinelles Lernen und
  natürlichsprachlicher Systeme
• Andreas Karwath

2
What is the seminar about?

• Applications of Data-Mining to real scientific or
  economic questions and problems
• Data-Mining
• Extracting implicit, previously unknown,
  potentially useful information from data
• We will investigate
• What is actual the problem/question
• ... and why can t it be solved manually
• What information can be used
• Problems with data representation
• What is the approach presented in the papers
• Have the authors solved the problem?
• Could the solution be improved with other
  approaches

3
Requirements

• 3 credit points
• Two presentations for each subject/paper theme
• one introductory (15 mins),
• one advanced (40 mins)
• Your presentation has to be send to me at least
  one week before in PDF (!) format.
• Report
• one written report (max. 14 pages Springer LCNS
  Style)
• deadline 31.08.2007
• Language
• English (slides! talk report)
• German (talk report)
• Previous knowledge/requirements
• Advantagous to have visited AI, MLDM, AAIT

4
Dates

• Block seminar
• first round
• most likely 06.06 (afternoon)
• second round
• most likely 20.06, 27.06, and 04.07 (afternoon
  5 talks)
• Up to date information
• http//www.informatik.uni-freiburg.de/ml/teaching
  .html
• -gt Sommersemester 2007 -gt Seminar -gt Further
  Information
• Office hours
• you have to arrange a meeting via email!
• Tuesdays/Thursday 0900 - 1000, 079, room 1012

5
Subject Allocation

• Send an email to karwath_at_informatik.uni-freiburg.d
  e by Wednesday 25.04 indicating
• A preference list of papers (at least three)
  using the paper ID (from web page)
• Your Name and MatriculationNumber
• Subject line DMtPS
• You will get an reply to tell you that I have
  received your email
• If you do not send an email I assume that you are
  not taking part in this seminar!!
• In case of preference clashes
• Ill try to mediate or
• I decide

6
Subjects

• Earth Science Climate Change
• Marketing Advertising
• Social Networks Human Behaviour
• Musicology
• Neuroscience
• Bioinformatics
• Medicine
• (Human) Activity Prediction

7
Musicology

• IDmus1
• Madsen S.T., Widmer G.
• Exploring Pianist Performance Styles with
  Evolutionary String Matching,
• International Journal on Artificial Intelligence
  Tools. World Scientific Publishing Company,
  15(4), 495-514. (2006).
• Keywords Piano playing, SOM, approximate string
  matching, evolutionary algorithms
• IDmus2
• Bergstra J., Casagrande N., Erhan D., Eck D.,
  Kégl B.
• Aggregate Features and AdaBoost for Music
  Classification.
• Keywords genre classification, artist
  recognition, audio feature aggregation, AdaBoost
• IDmus3
• Tobudic A., Widmer G.
• Relational IBL in Classical Music
• Machine Learning, 645-24 (2006)
• Keywords relational instance based learning,
  learning to play music,

8
Neuroscience

• Dneu1
• Fan Y., Shen D., Davatzikos C.
• Detecting Cognitive States from fMRI images by
  machine learning and multivariate classification
• MIUA 2006 PDF
• Keywords brain images, feature extraction SVM
• IDneu2
• Shenoy P., Rao R.
• Dynamic Bayes Networks for Brain-Computer
  Interfacing
• NIPS 2005, 17 PDF
• Keywords dynamic bayes networks, brain-computer
  interface (BCI), SVM

9
Activity/Intention Prediction

• IDact1
• Fogarty J., Au C. , Hudson S. E.
• Sensing from the basement a feasibility study of
  unobtrusive and low-cost home activity
  recognition
• UIST '06, 91-100, 2006.
• Keywords activity recognition, sensing in the
  home, sensor-based models, SVM, WEKA
• IDact2
• Shen J., Dietterich T.G.
• Active EM to reduce noise in activity recognition
• IUI '07, 132-140, 2007.
• Keywords active learning, expectation-maximizatio
  n (EM), intelligent interface, machine learning,
  noise
• IDact3
• Beetz M., v. Hoyningen-Huene N., Bandouch J.,
  Kirchlechner B., Gedikli S., Maldonado A.
• Camera-based observation of football games for
  analyzing multi-agent activities
• AAMAS '06, 42-49, 2006.
• Keywords analysis of intentional activity,
  motion interpretation, motion tracking, object
  tracking, state estimation, video analysis

10
Economics/Retail

• Deco1
• Cumby C., Fano A., Ghani R., Krema M.
• Predicting customer shopping lists from
  point-of-sale purchase data.
• KDD '04, 402-409, 2004.
• Keywords retail data mining, classification,
  machine learning (variety of algorithms)
• IDeco2
• Nikovski D., Kulev V
• Induction of compact decision trees for
  personalized recommendation
• SAC '06, 575-581, 2006
• Keywords frequent item-set mining, product
  recommendation, response modeling, decision trees

11
Biology/Bioinformatics

• IDbio1
• Huang C., Morcos F., Kanaan S.P., Wuchty S., Chen
  D.Z., Izaguirre J.A.
• Predicting Protein-Protein Interactions from
  Protein Domains Using a Set Cover Approach
• IEEE/ACM Trans. Comput. Biol. Bioinformatics,
  4(1), 78-87, 2007. PDF
• Keywords Computations on discrete structures,
  graph algorithms, bioinformatics (genome or
  protein) databases, biology, genetics.

12
Medicine

• IDmed1
• Ordonez C.
• Comparing association rules and decision trees
  for disease prediction
• HIKM '06, 17-24, 2006.
• Keywords association rule, decision tree,
  medical data
• IDmed2
• Rao R.B., Krishnan S., Niculescu R.S
• Data mining for improved cardiac care
• SIGKDD Explor. Newsl. 8(1), 1931-0145, 2006.
• Keywords Medical information systems,
  probabilistic reasoning
• IDmed3
• Wu H., Salzberg B, Sharp G.C., Jiang S.B.,
  Shirato H., Kaeli D.
• Subsequence matching on structured time series
  data
• SIGMOD '05, 682-693, 2005
• Keywords time series, clustering, tumor analysis

13
Geography/Weather/Climate

• IDgeo1
• Basak J. Sudarshan A., Trivedi D., Santhanam
  M.S.
• Weather Data Mining Using Independent Component
  Analysis
• J. Mach. Learn. Res., 5, 239-253, 2004
• Keywords priciple component analysis (PCA),
  weather data, spacio-temporal pattern mining

14
Networks

• IDnet1
• Bhattacharya I., Getoor L.
• Collective Entity Resolution in Relational Data
• TKDD Volume 1(1), 2007.
• Keywords citation analysis, entity resolution

15
Presentation Rounds

• First round (10-15 minutes)
• Your task Present the application area
• Problem specification
• Purpose
• Introduce the field to common students
• Check on presentation skills
• Get feedback for second presentation

16
Presentation Rounds

• Second round (35-40 minutes)
• Your task Present the application area (as a
  reminder) and the data-mining part
• Data generation, data extarction, feature
  construction
• Data mining technique used
• Results
• Have they solved the problem?
• Review what other solutions to this or related
  problems are around (!)
• Purpose
• Learn about data-mining
• Improve your presentation skills
• Take part in scientific discussions

17
Subject Allocation

• Send an email to karwath_at_informatik.uni-freiburg.d
  e by Wednesday 25.04 indicating
• A preference list of papers (at least three)
  using the paper ID (from web page)
• Your Name and MatriculationNumber
• Subject line DMtPS
• You will get an reply to tell you that I have
  received your email
• If you do not send an email I assume that you are
  not taking part in this seminar!!
• In case of preference clashes
• Ill try to mediate or
• I decide