Binary classification using parallel genetic algorithm

Binary classification is one of the most popular methods in supervised pattern classification. In this paper, we would like to propose an algorithm based on genetic algorithm for binary classification. Binary classification here is presented in a nonlinear programming form. Genetic algorithm is then used to search solutions of nonlinear programming. Four databases (one transcriptomics, one proteomics, and two breast cancers) were used to test the algorithm and six other well-known methods. Parallel computing based on island model was also experimented. The results show that the algorithm could identify most similar patterns in the database with high precision. Island model not only increases computational speed but also gives high quality result.