Abstract:

In order to overcome the disadvantages that LS-SVR algorithm is not suitable to
multiple input multiple output system modeling directly,a novel algorithm defined as
MLSSVR was proposed by adding sample absolute errors in objective function.And a novel flatness pattern recognition method
based on MLSSVR was put forward by applying MLSSVR algorithm on pattern
recognition.Then,comparison between the MLSSVR recognition method and the combination method of LS-SVR was conducted,and the recognition ability of MLSSVR recognition model was tested and analyzed. Experimental results demonstrate the validity of the MLSSVR algorithm. The flatness pattern recognition model based on MLSSVR can
avoid complex computation of LS-SVR combination method,enhance the recognition speed effectively,and  has higher recognition accuracy and good generalization ability.
 

Key words: least squares support vector regression(LS-SVR), multi-output least squares support vector regression(MLSSVR), flatness, pattern recognition

摘要：

为了克服最小二乘支持向量回归机(LS-SVR)算法不能直接应用于多输入多输出(MIMO)系统建模的缺点，通过在目标函数中加入样本绝对误差项，提出了一种多输出最小二乘支持向量回归机(MLSSVR)新算法。将MLSSVR算法应用于板形模式识别研究，提出了一种基于MLSSVR的板形模式识别新方法，将该方法与LS-SVR合成识别方法进行对比实验，并对MLSSVR识别模型的识别能力进行了测试和分析，
结果证明了MLSSVR算法的有效性。MLSSVR板形模式识别方法不仅避免了LS-SVR合成方法的复杂组合运算，具有更高的识别速度，而且具有更高精度和很强的泛化能力。


关键词: 最小二乘支持向量回归机, 多输出最小二乘支持向量回归机, 板形, 模式识别