Evolutionary multiobjective optimization for medical classification

We propose a computational environment based on evolutionary algorithm for medical classification. We use evolutionary multiobjective optimization (EMO) to solve a general medical minimization problem. As an example, we simultaneously minimize three objectives, namely the number of genes responsible for cancer classification while reducing the number of misclassifications in both testing and learning data sets for real patients. Results quality is reported against three genetic operators namely selection, crossover and mutation, each of which offering three different methods. Our implementation gives comparable results to more sophisticated methods, such as NGSAII-like ones, with far less computational efforts.