Classification Techniques

Classification Techniques

1. Divides large heterogeneous groups into smaller homogeneous

2. BENEFITS

   • increased certainty
   • Resource utilisation
   • Patient needs

3. USES

   • Medical Diagnosis
   • identifying risk groups
   • Predicting outcomes
   • Handling variability
   • capturing population behaviour

4. TECHNIQUES

   • Nearest neighbour
   • Discriminant Analysis
   • Cluster analysis
   • Regression
   • tree-based Algorithms
   • Neutral networks

5. CART

   • classification and regression trees
   • EXAMPLES
     • Harper et al. 2004 (maternity risk groups)
     • Harper et al. 2003 (diabetic retinopathy)
   • Algorithm
     • 1. total variance
     • 2. sum(x), Sum(x^2), N
     • 3. sort in increasing
     • 4. split to find minimum variance
     • 5. by: .
     • UNTIL: number of nodes or repeats OR no gain on variance