论文部分内容阅读
马田系统(MTS)是一种多元模式识别方法,它首先通过正常样本来建立基准空间,再利用正交表和信噪比来筛选有效变量,最后通过马氏距离来进行分类、诊断和预测.当建立基准空间的正常样本中掺杂少数异常点时,MTS的性能必然会受到影响.根据多变量控制图原理对建立基准空间样品的适合性进行判别,将在控制线外的样品点删除后建立新的基准空间,并通过UCI数据集进行可行性分析及分类效果比较,结果显示:经多变量控制图优化后的MTS,其性能得到显著提高.
The MTS is a multivariate pattern recognition method that first establishes the reference space through normal samples, then filters the valid variables by orthogonal table and signal-to-noise ratio, and finally classifies, diagnoses and predicts Mahalanobis distance The performance of MTS will inevitably be affected when a few abnormal points are established in the normal sample of the reference space.According to the multivariate control chart principle, the suitability of establishing the reference space sample is judged, and the sample points outside the control line are deleted After establishing a new reference space, the feasibility analysis and the comparison of the classification results by the UCI dataset show that the performance of the MTS optimized by multivariate control chart is significantly improved.