Numerosity Reduction
Choose alternative, smaller forms of data representation to reduce data volume. Basically removing rows/samples.
Methods
Parametric methods (regression, log-fitting)
- assume the data fits a model (e.g. a function)
- estimate model parameters
- store only parameters
- discard data
- e.x. log-linear model
Non-parametric methods:
- do not assume models
- Histogram Analysis
- Clustering
- Sampling
Data Cube aggregation:
- Aggregationstypen auf MDMs
- reference appropiate levels