Research And Implementation Of Milk Yield Prediction Of Dairy Cows Based On Ensemble Learning

Animal husbandry is an important branch of agriculture,and is of great significance in meeting people’s nutritional needs,promoting social development and raising the level of the national economy.Since China’s reform and opening up,the livestock industry has been developing rapidly,with the level of specialisation,industrialisation and organisation increasing significantly and the framework of the industrial system taking initial shape,while dairy farming,as an integral part of the livestock industry,has been listed by the state as an industry to be adjusted and encouraged for development,making China’s dairy farming industry achieve significant development.The physical appearance of a dairy cow is not only a direct reflection of body structure,but also has a close relationship with milk production,functional life and reproductive performance.Of these,milk yield is the most important indicator for assessing a cow’s production capacity and is a central factor in the economic efficiency of the dairy industry.The traditional method of predicting milk yield using cow body size uses a mathematical model of multiple regression equations,which is a single prediction method and does not quantify the error in the results.Most body size traits were discarded directly in the modelling and the data were not fully explored.Therefore,this paper takes Holstein cows provided by Ningxia H Dairy Company as the research object,collects data related to body size and conducts correlation analysis,and uses relevant techniques in the field of integrated learning to build a milk yield prediction model by using five existing more mature integrated learning algorithms,namely Random Forest,Adaboost,GBDT,XGBoost and LightGBM.To further improve the prediction accuracy of the model,three integration methods,Bagging,Boosting and Stacking,were used to improve the model,and a Bagging-XGBoost milk yield prediction model based on Stacking was proposed.The experimental results showed that the milk yield predicted by the model had less error with the actual one and was more superior compared with the model before improvement.In addition,as information technology empowers the dairy farming industry,Ningxia,as a national dairy development advantageous area,continues to promote large-scale operation,strengthen digital construction,and build an international first-class high-quality milk production base,but compared to developed regions,Ningxia dairy scale farming still has problems such as sloppy production process management,low level of intelligence,and difficulties in transforming results.Therefore,based on the optimisation of algorithms and models,this paper designs and implements an intelligent dairy farming management system by conducting a comprehensive demand analysis of the enterprise.An improved milk yield prediction model based on integrated learning is applied to provide technical support for intelligent management and decision making in the enterprise.