门限自回归模型及其在耕地总量预测中的应用

Threshold Auto-regressive Model and its Application in Gross Arable Land Prediction

利用梅州市55年来长时序耕地总量统计资料,引入门限自回归模型对耕地面积进行了预测,并和马尔可夫模型预测结果进行了比较。结果表明,通过门限值的控制作用,门限自回归模型有效地利用时序数据隐含的时序分段相依性这一重要信息,限制了模型误差,从而保证了预测性能的稳健性,有很高的短中期预测精度。

Long time sequence material of gross arable land in recent 55 years in Meizhou City was used to predict gross arable land by applying Threshold Auto-regressive （TAR） model. And its result was compared with that of Markov model. The result of the calculation showed that the scheme was practical and efficient. TAR model could effectively utilize the important information of the interdependence between section and time series dates, reduce model errors and ensure good stability and accuracy of the forecasting model.