基于样条逼近的评价方法研究

Study on evaluation method using spline approximation

样条逼近是从结构 (空间或时间 )性数据集合中提取特征的有效方法 ,本文提出基于样条逼近的评价方法 ,并进行评价方法应用研究 .首先给出可变结点的样条逼近中数据集合结点位置的选取准则 ,是使逼近的误差最小 ;给出数据集合结点位置的选取方法及算法 .对于上述分划的子区间 ,使用主成分分析方法对每个子区间进行评价 ,得到相应子区间的评价值 ,由于每个子区间的指标数不同 ,对子区间的评价值进行修正以体现这种差别 ;之后利用上述修正的子区间的评价值计算综合评价值 .

The spline approximation is an effective method that the features are abstracted in structural (space or time) data set. An evaluation method using spline approximation is proposed in this paper, and including study on its application. The criterion of selecting node location in data set using the spline approximation with variable nodes is introduced, as the criterion is to make the error least. The theory and algorithm of selecting node location in data set are presented. Every sub division is evaluated using principal component analysis method, and the evaluation values of the sub divisions are obtained. Because the index number in every sub division is different from other sub division, the evaluation values of the sub divisions are revised to express the difference. The comprehensive evaluation value is computed using revise evaluation values of the sub divisions. The application results of the method are presented.