32 Replacing a Mixture of Experts with a New GRNN Oracle as a Solution of the Complex Adaptive System for the Diagnosis of Breast Cancer

Problems involving complex adaptive systems (CASs) require complex adaptive solutions. In this work a new statistical mixture of experts architecture is presented as a solution to the general CAS problem. The new architecture is based on the general regression neural network (GRNN) foundation. This new mixture of experts solution is validated using a diagnostic mammography dataset. The outputs from various complex adaptive decision methodologies are used as inputs for GRNN based mixture of experts architecture. The work shows that this new solution outperforms the individual component models in making diagnostic benign-malignant predictions, which validates the solution.