Application of Variable Selection Method Based on Genetic Algorithm in Marine Enzyme Fermentation

Abstract

Genetic algorithm (GA), a global searching method, is applied to select the variables in soft sensing of marine enzyme fermentation. Compared with traditional methods of MIV and PCA, the optimal variables selected by GA have clear physical meaning and fewer numbers on the basis of variable selection frequency. After that, a soft sensing model based on BP neural network is established and a BP-GA soft sensing model is realized. Soft sensing results of enzyme activity show that, BP-GA model can provide better non-linear fitting ability and higher soft sensing accuracy, compared with the models of BP, BP-MIV and BP-PCA.