Generalization Based Knowledge Discovery
Attribute Oriented Induction
- Generalize the specific values of the task-relevant attributes to higher levels.
- Remove attributes when:
- further generalization is not possible OR
- the number of distinct values for an attribute at the higher levels exceeds the generalization threshold for that attribute.
- Merge identical tuples. Keep quantitative measure of tuples merged to allow quantitative presentation of knowledge acquired.
Two main algorithms:
- Spatial-Data-Dominant Generalization
- Non-Spatial-Dominant Generalization