A parallel genetic programming based intelligent miner for discovery of censored production rules with fuzzy hierarchy

Automated discovery of rules with exceptions and hierarchical structures is an important problem in data mining. A knowledge structure based on Censored Production Rules with Fuzzy Hierarchy (CPRFH) not only provides an excellent mechanism for handling exceptions but also captures the hierarchical relationship among the classes in the dataset. Moreover, CPRFHs are able to exhibit variable precision logic for approximate reasoning. This paper proposes discovery of knowledge in the form of CPRFHs using island model of genetic programming with two advanced genetic operators, namely; fission and fusion. The fission and fusion operators impart intelligence to the system as these operators discover new classes/concepts which are not present explicitly in the data set being mined. A suitable encoding with syntactic constraints is designed and an appropriate fitness function is suggested to measure the goodness of the hierarchies. The experimental results confirm that the island model with fission and fusion outperforms the sequential as well as the island models without fission and fusion in terms of correctness of the solution arrived and size of the trees evolved.