中文摘要：

对常用的回归方法进行研究.此类方法虽然几何解释明确、易于求解,但均须事先确定（或假定）变量间的因果关系,再考虑建模,在实际应用中,对于很难确定变量的因果关系的问题,如物联网数据分析,上述方法就会失效.为此,提出一种无需假定因变量的隐目标回归方法.该方法易于核化,可以推广到非线性回归问题.通过人工数据和国际标准数据集上的实验验证了所提算法的有效性.

英文摘要：

Commonly used regression methods were studied. Although the geometric interpretation of these methods is clear and the methods are easy to solve, a causal relationship must be determined （or assumed） in advance between variables, then the modeling is considered. In specific applications, such as the Internet of things data analysis where the causal relationship between variables is difficult to determine, the above method fails. To solve this problem, an implicit goal regression method without assuming the dependent variable in advance was proposed. This method is easy to nucleate and extendable to nonlinear regression problems. Experimental results on artificial data and international standard data sets verified the effectiveness of the proposed algorithm.