A parallel genetic programming for single class classification

Abstract

In this paper, we present an algorithm based on genetic programming for single (one) class classification that uses one set containing similar patterns in training process. This type of problem is called single (one) class classification, a novel detection. The proposed algorithm was tested and compared to seven other traditional methods based on two publicly available transcriptomic and proteomic time series datasets and two public breast cancer datasets. The results show that the algorithm could find most similar patterns in the databases with rather low misclassification rates. We also applied parallel genetic programming for this algorithm and it proves that the island model can give better solutions than sequential genetic programming.