Construction Of A Prediction Model For Dry Matter Intake And Milk Production Of Dairy Cows And Estimation Of Nutritional Requirements

In recent years,the scale of dairy farming in China has been continuously increasing,and the shortage of feed resources has become one of the limiting factors for the development of the dairy industry.Optimizing feed formulas to achieve the best feed conversion efficiency and achieving scientific feeding are important ways to alleviate the shortage of feed resources.Determining the optimal nutrient requirements for dairy cows is a basic approach for optimizing feed formulas and precision feeding.Machine learning can analyze and refine patterns from massive data,construct corresponding prediction models,and demonstrate strong predictive capabilities in multiple fields.This study took the indicators of body weight,parity,day in milk,and dietary nutrient composition obtained from actual pastures and literature as characteristic variables,and used machine learning algorithms to construct a prediction model for dry matter intake(DMI)and milk production of lactating cows.The requirements for crude protein,crude fat,starch,neutral detergent fiber,etc.are determined,providing a basic basis for optimizing the design of dietary formulas and improving feed utilization.Experiment 1:Constructing a prediction model for dry matter intake of dairy cows based on machine learning algorithmsThis study collected literature data and constructed a prediction model for cowDMI based on machine learning regression algorithm.By sorting out 119 literature,302 sets of effective data were collected,including cow age,body weight,parity,day in milk,lactation performance,and dietary nutrient composition,to construct a basic database for predicting DMI.Through data cleaning,CPM software calculation,feature engineering,screening of multiple mainstream regression algorithms,and parameter grid search cross validation,the GBR algorithm in the integrated regression algorithm was selected to construct a DMI prediction model.The independent variables included body weight,lactation days,milk yield,milk protein rate,milk fat rate,lactose rate,and the content of crude protein,starch,and crude fat in the diet,with a fitting coefficient R~2of 0.775.Comparing this model with the prediction model in the nutritional needs of cow NRC(2001),it was found that the prediction models R~2,MAE,MAPE,RMSE,and other indicators were better than the NRC(2001)model,indicating that the model constructed in this experiment can more accurately predict cow DMI.Experiment 2:Constructing a milk production prediction model for cows based on machine learning algorithms and determining nutritional requirementsThis experiment collects actual data from pastures,constructs a prediction model for milk production of cows based on machine learning regression algorithm,and determines nutritional requirements.The experiment obtained 11562 sample data from Jun Le Bao Dairy Farm.The DMI of cows was calculated using the DMI prediction model in Experiment 1.Through data integration,preprocessing,feature engineering,preliminary screening of regression algorithms,and cross validation of parameter grid search,the XGB regression algorithm was ultimately determined to be used as the model construction algorithm based on the prediction performance of each algorithm.The independent variables included body weight,day in milk,rumination time,parity,and crude protein.The fitting coefficient R~2of 8 characteristic indicators,including starch,neutral detergent fiber,and daily intake of crude fat,reached 0.856.Based on this model for machine learning visualization operations,and by controlling independent variables or combining the intake of crude protein,starch,neutral detergent fiber,and crude fat in pairs,daily milk production could be predicted,and nutrient requirements for different milk production levels could be inferred from the predicted milk production.This study successfully constructed a prediction model for DMI and milk production in dairy cows,with a prediction effect of R~2exceeding 0.75,indicating relatively accurate prediction results.And the requirements of dairy cows for nutrients such as crude protein,crude fat,starch,and neutral detergent fiber were obtained,which made a beneficial attempt to estimate nutrient requirements using big data in actual production and provided a reference basis for the design of precise feed formulas for dairy cows.