Prediction of frozen semen doses production in dairy studs using machine learning algorithm

The ability to predict the frozen semen doses produced per ejaculate would be of considerable benefit for the management of skill, human resource, capital and time. The new computing paradigm called machine learning involves in predicting dependent variable by learning complex and non-linear relationship among independent variables. The purpose of this study is to develop prediction model using one of the conventional and machine learning modelling techniques called Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) respectively. A Total of 1,57,532 ejaculates data were used for modelling. The modelling involved prediction of frozen semen doses produced per ejaculate using independent variables namely volume of ejaculate, ejaculate number, sperm concentration, initial motility and post thaw motility. Various combinations of architectural parameters were employed to explore optimum configuration for each model. The ANN (R2=90.66) modelling was observed to be efficient over MLR (R2=73.52). The root mean squared error (RMSE) value was found to be lower in ANN (33.89) when compared to MLR (57.31). Hence, the ANN modelling approach is efficient to predict frozen semen doses that could be produced per ejaculate.