Top 10 Data Mining Algorithms
8. k Nearest Neighbors
- Majority Voting vs. Weighted Majority Voting
- What should k be?
- What should the similarity metric be? Euclidian? Cosine?
- With certain assumptions, error is bounded above by twice
the Bayes error.
- Condensation and editing can both be used to enhance performance.
9. Naive Bayes
- Can be easily modified to handle continuous values.
10. CART
- CART: Classification and Regression Trees appeared in 1984.
- Creates binary trees.
- CART contains many insightful ideas, but C4.5 is the more
important decision tree to know.
Machine Learning
