Modelling the Chemical and Microbial Changes of Saffron Flower during Storage Using Artificial Neural Networks and Genetic Algorithm

Saffron, as the most expensive agricultural and pharmaceutical product of the world, has a special value among plants. Since the Saffron harvesting period is short, its storage for later processing requires understanding the most effective factors affecting the quality of saffron and its deterioration. Therefore the effects of reposition thickness, storage temperature and storage time of saffron flowers on its chemical parameters including crocin, safranal and picrocrocin values of saffron stigma and its microbial quality indicators including total count, coliform and mold contamination were modelled. This was done using multi-layer perceptron artificial neural network (ANN) and its structure and the learning parameters were optimized using genetic algorithm technique. The optimized MLP neural networkwas capable to predict the saffron quality characteristics during storage with coefficient of determinations higher than %94 and low error values (RMSE lower than 3.5 for all responses). The ANN model showed that reposition thickness has the lowest impact on chemical and microbial parameters deterioration while increasing storage temperature and time drastically increased loss of quality although the effect of storage time is lower than that of storage temperature. Overall, keeping fresh saffron flowers at a low temperature near zero degrees centigrade is necessary for maximum retention of valuable chemical compounds and minimum microbial contamination development during saffron flower storage for further processing.