Research On Yili Horse Weight Estimation System

The weight of the horse is an important indicator that could reflect and measure horses'health status.It has important reference significance in horse breeding,meat quality evaluation,feeding management,and horse identification,etc.However,Yili horses have giant shape,a lively temperament and in the wild status for a long period,which makes it hard to be weighed.This paper mainly researched the performance of several machine learning models in the estimation of Yili horse weight and presented those models to users in the form of webpages.The main data of this research is got Zhaosu County,Yili Area,Xinjiang,which is one of the central breeding areas of Yili Horse.The research objects are two-year-old Yili horse,and information Such as age,gender,chest circumference,body length,height,circumference of cannon bone and weight is collected.By analyzing and screening the data,the chest circumference,body length,height,circumference of cannon bone and weight are selected as the eigenvalues of the model.Model are selected based on machine learning.First,the selection and establishment of common linear regression models are selected and then come to the uncommon linear regression models.Then,by using the idea of ensemble learning,individual models are aggregated.Finally,a neural network model is tried to be established to solve the problem.The development of the system follows software engineering specifications.First the analysis and design of the system in the early stage,then the development and debugging of the system in the middle stage,and finally are the system is tested and operated in the final stage.The research result can be concluded in the following five parts.(1)Common machine learning linear regression models are studied in the weight estimation model of Yili horses,which includes the least-squares linear regression model,ridge regression model,Lasso regression model,and elastic network model.(2)Uncommon machine learning linear regression models are also constructed,including the support vector regression,regression decision tree,K nearest-neighbor regression.(3)Construction of the simple average integration model and stack-level integration model based on the idea of ensemble learning.(4)Research of the radial basis function(BRF)in a neural network model in machine learning.(5)The weight estimation system contains the functions of displaying four types of characters of tourists,logged-in users,normal administrators and system administrators on the homepage,as well as the displaying of news and announcement,weight estimation,weight model training,personnel management,function management,and background management,etc.After analysis,the research results indicate that in the model for estimating the bodyweight of Yili horse,the collinearity problems exist between independent variables,which can influence the estimation accuracy of the least squares regression model.These collinearity problems of the independent variables can be solved by the ridge regression model to a certain degree,and thus improve the goodness of fit of the model.With few independent variables,the lasso regression model and the elastic network regression model fail to eliminate the collinearity problem completely.In the case of small sample,the support vector regression model with a linear kernel function has bad linear fitting result.The linear fitting of the stacked layered integration model with sophisticated structures is even worse than the simple averaged integration model.Among each training model,including the common horse weight estimation model,the neural network model with radial basis function(RBF)has the best linear fitting goodness,which indicates the collinearity problem effectively in the case of small sample can be eliminated by the model.