Zero Rule or ZeroR is the benchmark procedure for classification algorithms whose output is simply the most frequently occurring classification in a set of data. If 65% of data items have that classification, ZeroR would presume that all data items have it and would be right 65% of the time.
ZeroR is a simple and effective benchmark: if an algorithm correctly predicts classifications less frequently than ZeroR, it is obviously of no value for the domain in question!
Compare One Rule.
- alias
- ZeroR
- subtype
- has functional building block
- FBB_Classification
- has input data type
- has internal model
- has output data type
- ODT_Classification
- has learning style
- LST_Supervised
- has parametricity
- PRM_Nonparametric
- has relevance
- REL_Benchmark
- uses
- sometimes supports
- mathematically similar to