RECURSIVE GENERALIZED M ESTIMATES FOR AUTOREGRESSIVE MOVING-AVERAGE MODELS

Abstract. Outliers in time series seriously affect conventional parameter estimates. In this paper a robust recursive estimation procedure for the parameters of auto-regressve moving-average models with additive outliers is proposed. Using 'cleaned' residuals from an initial robust fit of an autoregression of high order as input, bounded influence regression is applied recursively. The proposal follows certain ideas of Hannan and Rissanen, who suggested a three-stage procedure for order and parameter estimation in a conventional setting.
A Monte Carlo study is performed to investigate the robustness properties of the proposed class of estimates and to compare them with various other suggestions, including least squares, M estimates, residual autocovariance and truncated residual autocovariance estimates. The results show that the recursive generalized M estimates compare favourably with them. Finally, possible modifications to master even vigourous situations are suggested.     

Application of a rice field experimental error distribution function to nitrogen-phosphorus-potassium fertilizer response model analysis

A distribution function of rice yield deviations from the mean was developed from field experiments with 555 plots at 16 sites in Zhejiang province, China, for three years. The deviation distribution in interval of 50kg/ha appeared as a symmetrical distribution with a high peak (Mean=0.279 [kg/ha], SD=240.686 [kg/ha]). Normality test using Kolmogrove-Smirnov test between the observed cumulative distribution and the normal cumulative distribution function indicates that the observed deviation distribution is not normal. An empirical exponential cumulative distribution function was developed. The distribution function was used to remove outliers during the development of a rice yield fertilizer response model, based on data from a non-replicated NPK field experiment.     

一种基于自然最近邻的离群检测算法

任何涉及k近邻求解问题的算法被应用于处理不同特征的数据集时,参数k值的选择都会明显影响算法的性能和结果。因而,如何选择k近邻算法中敏感参数k值一直是一个研究难点。提出了一种新的近邻关系———自然最近邻,它不需要设置参数k,每个节点的邻居是由算法自适应计算而形成的。针对离群点检测的特殊性,通过确定自然最近邻居搜索算法的终止条件,提出一种基于自然最近邻的新的离群检测算法ODb3N。实验表明,该算法不仅避免了k近邻中参数的选择问题,而且能够更有效地发现离群簇。     

基于井下人员的异常轨迹分析方法研究

针对井下作业人员轨迹数据信息多维度和稀疏性等问题,提出了基于离群点的异常轨迹筛选ZFMTRAOD算法,首先通过对轨迹子段建立R-tree索引提升检索速度,然后利用离群检测思想对邻域半径内轨迹子段的数量和平均时间判断轨迹是否异常,最后利用井下作业人员的轨迹数据对算法的性能进行比较,发现基于离群点的井下人员轨迹分析算法不仅能判别出井下作业人员异常轨迹的类型,还提高了异常轨迹判别的准确率。     

Mixtures of robust probabilistic principal component analyzers

Mixtures of probabilistic principal component analyzers model high-dimensional nonlinear data by combining local linear models. Each mixture component is specifically designed to extract the local principal orientations in the data. An important issue with this generative model is its sensitivity to data lying off the low-dimensional manifold. In order to address this problem, the mixtures of robust probabilistic principal component analyzers are introduced. They take care of atypical points by means of a long tail distribution, the Student-t. It is shown that the resulting mixture model is an extension of the mixture of Gaussians, suitable for both robust clustering and dimensionality reduction. Finally, we briefly discuss how to construct a robust version of the closely related mixture of factor analyzers.     

基于单元的孤立点检测算法改进及应用

研究基于单元的孤立点检测算法,给出数据空间的单元格划分及数据对象分配算法。针对该算法中阈值M设置的不足,对算法进行改进并应用于纳税行为的分析。与其他孤立点检测算法对比的结果表明,该算法不仅能有效挖掘纳税行为中的孤立点,还能确定孤立点的位置,有利于对纳税行为的分析。     

基于最近邻相似度的孤立点检测及半监督聚类算法

传统的聚类算法是一种无监督的学习过程,聚类的精度受到相似性度量方式以及数据集中孤立点的影响,并且算法也没有很好的利用先验知识,无法体现用户的需求。因此提出了基于共享最近邻的孤立点检测及半监督聚类算法。该算法采用共享最近邻为相似度,根据数据点的最近邻居数目来判断是否为孤立点,并在删除孤立点的数据集上进行半监督聚类。在半监督聚类过程中加入了经过扩展的先验知识,同时根据图形分割原理对数据集进行聚类。文中使用真实的数据集进行仿真,其仿真结果表明,本文所提出的算法能有效的检测出孤立点,并具有很好的聚类效果。     

基于最小二乘支持向量机算法的测量数据时序异常检测方法

将最小二乘支持向量机方法引入火电厂DCS的测量数据时序异常检测领域,该方法很好地建立了火电厂DCS的测量数据时序预测模型,具有预测真实值能力强、全局优化及泛化性好等优点。将该方法应用于某600 MW超临界火电机组DCS测量数据中,经过训练后的LS-SVM模型对再热蒸汽温度数据的检验样本进行不良值检测与真实值预测,均方根误差和平均相对误差分别为0.067%和0.050%,均方根误差是BP网络模型、RBF网络模型的8.756%和8.272%,平均相对误差是BP网络模型、RBF网络模型的7.541%和7.236%。应用结果表明,最小二乘支持向量机方法优于多层BP与RBF神经网络法,能很好地满足异常检测与真实值预测要求。     

Implementation of a crack propagation constraint within a structural optimization software

A crack propagation constraint related to the stress intensity factor is examined for the minimum-weight design optimization of a composite blade-stiffened panel. A low-fidelity approach uses a closed-form solution for the stress intensity factor, while a high-fidelity approach uses the stress distribution around the crack. Structural optimizations are performed by low- and high-fidelity approaches for a number of panels configured with different values of the load, crack length, and blade height. Response surface (RS) approximations are then constructed for the optimal weight as a function of the three configuration variables. The computational cost, numerical noise, and accuracy for the two approaches are compared. An additional constraint in the low-fidelity solutions is found to be active for some of the configurations, increasing the difference between the low-fidelity and high-fidelity optima. It is shown that outlier analysis helps to identify the configurations with the largest difference.     

观测值中野值的判别与处理方法仿真研究

在Kalman滤波应用过程中,观测值中的野值是影响滤波效果的重要因素。分析了野值对Kalman滤波的影响机理,从滤波算法的原理和工程实践的实际出发,提出了一组野值判别和处理方法。仿真计算和工程实际应用表明,该方法性能可靠,简单易行,可以有效地消除野值对滤波的不理影响,提高滤波的精度。