CONTENTS

1. Distance metric definition
2. Ideal generalization threshold
3. DISTORTION: Equivalence class wise quality metric, Attribute wise quality metric, Overall quality metric
4. HEIGHT-COST-DISTANCE scheme

PRE-PREQUISITES

• Fix priorities manually.
• Calculate weights pertaining to priorities assigned.
• modification rates (preliminary analysis only): percentage of quasi ID cells changed. Does not include level to which generalisation occurs.

The metric scheme takes into account two types of attributes - numerical and categorical. Numerical attributes have a hierarchy based on their absolute values. Categorical attributes have a hierarchy that can be represented using a tree structure.

1. Distance Metric Definition

with normalisation:

Numerical -

N represents an Attribute, i and j represent different values of the attribute

Categorical -

without normalisation:

Numerical -