Evolutionary q-Gaussian Radial Basis Function Neural Network to determine the microbial growth/no growth interface of Staphylococcus aureus

In this paper, q-Gaussian Radial Basis Functions are presented as an alternative to Gaussian Radial Basis Function. This model is based on q-Gaussian distribution, which parametrizes the Gaussian distribution by adding a new parameter q. The q-Gaussian Radial Basis Function allows different Radial Basis Functions to be represented by updating the new parameter q. For example, when the q-Gaussian function takes a value of q→1, it represents the standard Gaussian Radial Basis Function. The model parameters are optimized through aMemetic Algorithm that evolves both its structure and connections. To evaluate the effectiveness of the model, it is tested with a real problem of predictive microbiology. The problem consists of determining the growth boundaries of Staphylococcus aureus, a food borne pathogen responechnique (SMOTE) -Gaussian Radial Basis Function Neural etwork redictive Microbiology emetic Algorithm sible for several outbreaks. The data from the study of [1] belongs to growth/no growth conditions of S. aureus whose temperature, pH and water activity (aw) has been divided into three categorical classes: growth (G), growth transition (GT) and no growth (NG). Due to the imbalanced nature of the problem, it has been necessary to apply an over-sampling algorithm. The over-sampling procedure selected was the Synthetic Minority Over-Sampling Technique (SMOTE) algorithm. This algorithm has been applied to the patterns in the minority class in order for the performance of the classifier in this class to be acceptable (the minority class in this problem is of vital interest).