基于豪斯托夫距离和二次惩罚支持向量机的鲁棒区间回归研究

Robust Interval Regression Based on Haustoff Distance and Quadratic Penalty Support Vector Machine

二次惩罚支持向量机区间回归模型从外部和内部两个方向逼近区间数据,能够较好地估计模糊现象中存在的内生不确定性。然而二次惩罚支持向量机区间回归模型的回归性能易受噪声数据的影响,因此本文采用豪斯托夫距离(Hausdorff距离)和豪斯托夫距离作为衡量区间数据的距离标准,通过K-近邻(KNN)算法对噪声数据进行有效筛选,从而提高噪声环境里的二次惩罚支持向量机对区间数据的回归性能。

The interval regression model of quadratic penalty support vector machine approximates the interval data from the external and internal directions, which can better estimate the internal uncertainty in the fuzzy phenomenon. However, the regression performance of the interval regression model is easily affected by the noise data. Therefore, this paper uses the Hausdorff distance as the distance standard to measure the interval data, and selects the noise data by KNN algorithm, so as to improve the regression performance of the interval data in the noise environment.