The determination of somatic cell count by fully connected artificial neural networks with feed forward back propagation multilayer perceptron

Within the scope of the thesis presented, it is aimed to generate an alternative model to classical methods in the estimation of the somatic cell count (SCC) is the gold standard indicator of subclinical mastitis using artificial neural networks (ANN) model which is a machine learning method from the physical properties of milk such as electrical conductivity (EC), pH, density and temperature at fore milking (TFM). Model performance was evaluated by testing with cross validation on a data that was not previously introduced to the model before and compared with the classical linear model (multiple linear regression) defined as the control model. The results showed that the model has satisfactory results in terms of loss and achievement performance (AIC=-338; HKOK=0.01; R2=0.95). Despite using a high number of parameters (81) comparison to the control model, it was observed that the performance of the test model (AIC=-338) created with ANN was higher than the control model (AIC=-240) created with the classical linear model. More successful prediction performance can be obtained as a result of the learning to be carried out using big-data obtained from automated milking information (such as milk yield, lactation sequence and period) on the machines called super-computer or parallel systems.