Approaches to Classification  

• Density estimation coupled with a decision rule
• Define a metric space and decide based on proximity (e.g K-nearest neighbour)
• project the attribute space into decision regions, e.g.
• Logistic regression: log-linear approximation
• discriminant analysis: linear separation
• decision trees: piecewise constant approximation
• perceptrons: linear separators
• neural nets: non-linear sparators

---