基于ARIMA模型对宁夏地区奶牛体细胞数的趋势预测

本研究通过对生产性能测定（DHI）数据的挖掘与分析建立预测宁夏地区奶牛体细胞数（SCC）的模型，为奶牛乳房炎的防制提供借鉴，使得DHI数据更加有效、及时地指导奶业的发展。对2011年9月~2016年2月宁夏地区奶牛平均SCC数据进行差分使其达到平稳化，采用季节ARIMA模型对数据进行分析、拟合和预测。利用R软件的auto.arima函数计算出合适的时间序列模型ARIMA（1，1，0）（1，1，0）_[12]，其AIC为-3.67。Acf检验说明残差没有明显的自相关性；Ljung-Box测试显示所有的P值>0.5，表明残差为白噪声，说明此模型可用来对未来的24个月进行预测。再利用R软件的forecast函数对2016年3月~2017年2月的数据进行预测，作出预测图。从预测的结果可以看出，宁夏地区奶牛SCC整体呈现下降趋势。2017年1月SCC最少，预测值约为25.31万个/mL；2016年3月SCC最大，预测值约为43.96万个/mL。从结果也可看出，宁夏地区奶牛SCC均大于隐性乳房炎的临界值（>20万个/mL），说明宁夏地区还应该加大对奶牛乳房炎的防制。同时，若能及时添加新的SCC数据，就能对该数据模型进行更新，使其预测值更接近真实值，对实际生产的指导意义更大。

Predicting the Dairy Cow Somatic Cell Number of Ningxia Area Based on ARIMA Model

Through dairy herd improvement (DHI) data analysis to predict somatic cell count (SCC) of Ningxia area in time, which providing a reference for the prevention and treatment of mastitis in dairy cows, making the DHI data more effective and timely in guiding the dairy industry production. Using the difference method to make the average cow somatic cell data form September 2011 to February 2016 stabled, then used the seasonal ARIMA model to analysis, fitting and forecasting data. The auto.arima function of R software had been used to calculate the optimal time series that finally confirmed the model was ARIMA (1,1,0)(1,1,0)_[12], AIC was -3.67. The Acf test showed that the residual had no significant autocorrelation; Ljung-Box test showed that all P-value>0.5, indicating that the residual was white noise, and this model could be used to make predication for the next 24 months. The forecast function of the R software was used to predict the dairy cows SCC from March 2016 to February 2017, and the forecast map was drawn. The predicted results showed that the SCC of the entire Ningxia area were showing a downward trend. The SCC would be the least in January 2017,and the predictive value was about 253 100 per mL. It would be the largest in March 2016, was about 439 600 per mL. The results also showed that the dairy cows SCC in Ningxia was higher than the critical value of 20 million of subclinical mastitis. It suggested that the prevention and treatment of dairy cow mastitis need to be strengthened in Ningxia area. At the same time, if the data of the dairy cows SCC was added in timely, the data model should be updated to make it more close to the true value, which would be more meaningful to the actual instruction.