中文摘要：

机器学习回归方法被广泛应用于复杂工业过程的软测量建模，κ-最近邻（κNN）算法是一种浒的学习算法，可用于函数回归问题，然而，传统κNN算法存在运行效率低，距离计算忽略牲权值的缺点，本文引入了二次型距离定义和样本集剪辑算法，改进了传统κNN回归算法，并将改进的算法用于工业过程软测量建模，仿真实验得到了一些有益的结论。

英文摘要：

Recently, machine learning regression algorithms are widely applied to soft sensor modeling for complex industrial processes. The κ-nearest neighbor （κNN） algorithm is a popular learning algorithm for solving regression problems. However, the traditional κNN algorithm has low efficiency and ignores the feature weights in distance computing. Using a quadratic distance definition and a data set editing algorithm, we have modified the traditional κNN regression algorithm. The modified algorithm is applied to soft sensor modeling and some useful conclusions are reached.