1
Data Mining on a Mushroom Database

• Clara Eusebi,
• Cosmin Gilga,
• Deepa John,
• Andre Maisonave

2
Presentation Summary

• Background Concepts
• Literature Review
• Focus of Study
• Research Methodology
• Results of Study
• Mushroom Database Application
• Future Research
• Conclusions

3
Background

• Algorithms and Techniques
• Jeff Schlimmers Dissertation
• Confusion Matrix

a b
500 0 a e edible
5 495 b p poisonous
4
Literature Review

• Overview of Data Mining
• Decision Trees
• Visual Classification and Human-Machine
  Interaction

5
Focus of the Study

• Run algorithms in Weka on the Mushroom Database
• Mushroom Database Application
• Edible or Poisonous?

6
Research Methodology

• Unpruned decision tree
• Classifiers that do not generate rules
• A classifier that does generate rules
• Jeff Schlimmers optimal rule set

7
Results of the Study

• Results on client databases
• much higher accuracy
• Results on Schlimmers database
• Elaborate unpruned tree

8
Mushroom Database Application

• Available at
• http//utopia.csis.pace.edu/cs691/2007-2008/team6
  /Mushroom_Database_Application.html

9
Future Research

• Differences between Dr. Chas data and
  Schlimmers data

10
Conclusions

• J48 Unpruned Tree
• Highest accuracy results
• Mushroom Database Application
• Human-Machine Interaction