自适应模糊ARDL时间序列半参数模型
An Adaptive Fuzzy Autoregressive Distributed Lag Time Series Based on Semi-Parametric Model

传统的模糊最小二乘时间序列模型在包含异常值的模糊数据集中表现出较差的性能。通过引入异常值检测策略，提出了一种降低异常值对未来预测影响的方法。为此，本文介绍了一种基于半参数技术的自适应模糊自回归分布滞后(ARDL)时间序列模型，该模型将半参数技术与加权最小二乘结合，根据加权平方距离误差，提出了在存在异常值时确定精确系数的估计程序，然后利用迭代算法对模糊时间序列模型的参数进行了估计。本文还扩展了传统时间序列模型中常用的几种拟合优度准则，以比较所提出的模糊时间序列方法与现有方法的性能。通过实际例子，验证了本文方法的有效性。结果清楚地表明，在模糊数据存在潜在异常值的情况下，本文提出的模型对模糊时间序列数据的预测具有可行性和有效性，优于其他模糊时间序列模型。
The traditional fuzzy least squares time series model shows poor performance in fuzzy data sets containing outliers. By introducing the outlier detection strategy, a method is proposed to reduce the influence of outlier on future prediction. For this purpose, this paper introduces an adaptive fuzzy autoregressive distributed lagdistributed (ARDL) time series model based on semi-parametric techniques. The model combines the semi-parametric techniques with the weighted least squares and proposes an estimation procedure to determine the exact coefficient when there are outliers according to the weighted square distance error. Then, the parameters of the fuzzy time series model are estimated by an iterative algorithm. Several goodness-of-fit criteria commonly used in traditional time series models are extended to compare the performance of the proposed fuzzy time series method with existing methods. A practical example is given to verify the effectiveness of this method. The results clearly show that the proposed model has reliability and effectiveness for the prediction of fuzzy time series data when there are potential outliers in fuzzy data, which is better than other fuzzy prediction models.