Optimizing Fuzzy Clinical Decision Support Rules Using Genetic Algorithms

In this paper, we present a technique for optimizing a fuzzy system using a genetic algorithm that works for patient status monitoring in the operating room. The genetic algorithm adjusts rule weights, outputs, and input membership functions to maximize the area under a receiver operator curve (ROC) for final classification. Compared to pre-optimization, the optimized fuzzy inference system increased ROC area from 0.68 to 0.77, which can be translated to an increase in specificity from 74% to 82%, at a fixed sensitivity of 58%