Modeling of Different Concentrations of Phytoestrogens Effect on Progesterone Hormone Levels in Farmed Female Huso huso Using Artificial Neural Network

The present study aimed to model different concentration of genistein and equol onfluctuations of 17ɑ - hydroxyprogesterone hormone levels in reproductive growthduring a year. After statistical analysis of the results obtained from experiments on54 female farmed Huso huso, the on effects of 0.2, 0.4, 0.8 and 1.6 g/kg of genisteinand equol on progesterone levels were used to train multi-layer perceptrone feedforwardartificial neural networks. Multi-layer perceptrone (MLP) with post errorexpression learning algorithm (momentum learning functions) and sigmoid activatorfunctions were used for modeling of networks in Neurosolutions version 6. Forevaluating of neural network models, regression (R2) and mean square error index(MSE) were used. The input parameters are phytoestrogens concentrations andseasons and the output parameter was progesterone level. Various applied networksgenerated easily associations of plasma progesterone and gonad development,provided a powerful tool for estimation. The accuracy of the trained network wasexamined with data from fish. Based on results, neural network system predictedprogesterone levels with high performance r = 0.93 and MSE = 0.00013. Therefore,it is possible to model the effect of phytoestrogens on progesterone with highcorrelation according to actual data in this species and also other species, which maybe useful for decreasing the costs of research and times.

Keywords:

Modeling , Artificial neural network , Progesterone , Huso huso