BP神经网络在中国西门塔尔牛屠宰性状早期预测中的应用

Application of BP Neural Network to Predict Slaughter Traits in Chinese Simmental

本研究运用DPS(Data processing system数据处理系统)软件构建BP神经网络,选取156头中国西门塔尔牛屠宰前7个重要生产性状(平均日增重、宰前活重、体高、体长、胸围、腹围、管围)来预测中国西门塔尔牛胴体重和屠宰率两个屠宰性状,得到了7-5-2 BP神经网络模型。通过对预测结果与实测结果进行统计分析从而对所构建的BP神经网络的有效性进行验证。仿真测试结果表明:胴体重和屠宰率的预测结果与实测结果的相关系数(R)分别达到0.8264和0.8967,胴体重和屠宰率预测相对误差分别为5.0%和2.1%,验证了该BP神经网络模型在中国西门塔尔牛屠宰性状早期预测中的可行性和实用性,对加快中国西门塔尔牛的选育进程,降低屠宰性状的测定成本具有一定意义。

In this study, DPS（ Data processing system） software was used to build BP neural network ,7 ante-mortem important traits （ ADG, ante-mortem live weight, body height, body length, chest circumference, abdominal circumference, cannon bone circumference） of Chinese Simmental were selected to predict the carcass weight and dressing percentage, and a 7-5-2 BP neural network model was obtained. By comparison of the predicted and measured results, the validity of the BP neural network constructed was verified. Simulation test results showed that： the correlation coefficient （r） of carcass weight and dressing percentage between the predicted and measured results was 0. 8264, 0. 8967, and the relative prediction error of carcass weight and dressing was 5.0 %, 2.1%, respectively, which proved the feasibility and practicality of the BP neural network model in Chinese Simmental early prediction of the slaughter traits. It was meaningful for speeding up the process of Chinese Simmental breeding and reducing the cost of slaughter traits.