线性模型与机器学习模型在牦牛体重预测比较

本研究旨在比较线性回归模型与机器学习模型在利用体尺性状预测体重时的准确性。测定102头革吉那布地区两岁龄牦牛相关体尺性状(体高、体长和胸围)与体重,然后将数据按照不同比例梯度分为训练集和测试集,利用传统的一般线性模型方法和机器学习方法(高斯过程回归、支持向量机)分别构建体尺性状与体重之间的回归预测模型。每个比例均重复5次,将体重的真实值与预测值之间的相关系数均值作为当前比例下的模型准确性结果。结果显示,随着训练集数据的增加,线性回归模型的预测结果较稳定在0.71至0.80之间,而机器学习方法的预测准确性最高达到0.91。故在训练集数据充足的情况下,相比于一般线性模型,利用机器学习方法进行预测具有更高的准确性。

Comparison of the Accuracy of Linear Model and Machine Learning Model in predication of Yak Body Weight

The aim of this study was to compare the accuracy of linear regression models with machine learning models in predicting body weight using body size traits. The body size traits (body height, body length and chest circumference) and body weight of the two-year-old calves in the heads of the Ginnabu area were determined. The data were then divided into training sets and test sets according to different proportions, using traditional general linear model methods and machines. The learning method (Gaussian process regression, support vector machine) constructs a regression prediction model between body size traits and body weight. Each ratio was repeated 5 times, and the correlation coefficient between the true and predicted body weights was used as the model accuracy result at the current scale. The study found that with the increase of the training set data, the prediction result of the linear regression model is stable between 0.71 and 0.80, and the prediction accuracy of the machine learning method is obviously improved, up to 0.91, so compared with the general linear model. Has a significant improvement.