Reasearch On The Application Of Data Mining Technology In Intelligence Animal Husbandry

Animal husbandry is the pillar industry in Sanjiangyuan area,and yak breeding is an indispensable part of industrial development.With the expansion of breeding scale and the development of technology,yak breeding and research have made great achievements.However,it still faces problems such as high cost,poor profit and wrong decision.This paper focused on the study of yak weight estimation,classification and growth model establishment,collected relevant data from some ranches in Qinghai,and analyzed by methods of data mining and artificial intelligence.So that the breeding and decision-making personnel can make scientific decisions in the yak breeding feed proportion,fattening and slaughter,reproduction breeding,disease early warning,product monitoring and etc.process.The main findings are as follows:(1).Multiple linear regression(LR),support vector machines(SVMs)and random forest(RF)were used to estimate the weight of yak by using simple size dimension,which take yak body height(BH),body length(BL),chest girth(CG)and circumference of cannon bone(CCB)as input variables and body weight as response variables.And the predicted accuracy of LR algorithm is 0.997,RF algorithm is 0.998,and SVMs algorithm is 0.820.The RF and LR algorithms showed great predictive ability in the yak dataset,and the error was within acceptable range(generally considered to be less or equal to 10%).(2)In order to solve the problem of selection and breeding optimization of yak,the classification of yak was predicted by using three methods : multiple Logistic regression,typical discriminant function and multi-layer perceptron artificial neural network,which according to the growth body attributes,scoring data and manual grading data of yak to carry out the classification.The artificial neural network method based on multi-layer perceptron has the best classification performance and the prediction accuracy reaches 97.7%.The prediction accuracy of the multiple Logistic regression algorithm and the discriminant function is 96.5% and 95.1% respectively.The results showed that the artificial neural network(Ann)method based on multi-layer perceptron can classify the yak grades scientifically and effectively.(3)In order to better describe the growth and development rules of yak,Compertz,Logistic,Von Bertalanffy and Brody growth models were respectively used to fit the body size and body weight of yak.It was found that the Von Bertalanffy model had the best fitting effect on yak weight,Logistic model had the best fitting effect in the analysis of yak circumference of cannon bone.However,Brody model showed good fitting performance in the analysis of body height,body length and chest girth of yak.Through the loop back test model,the average relative error between the measured value of the Von Bertalanffy model and the original mean value was 8.801%.The mean relative errors of the Logistic model for the measurement of tube circumference were 2.146%,while the mean relative errors of the Brody model for the measurement of body height,body oblique length and chest circumference were 4.280%,2.367% and 0.175%,respectively.The results showed that the model can fit the growth and development of yak well.The above research results can provide strongly support from data for decision makers to make scientific decisions in the processes of weight estimation,grade classification,growth and breeding management of yak,which is of great significance to yak farming.