论文部分内容阅读
针对产品销售时序包含噪声的数据特征,提出一种基于自适应分段损失函数的支持向量机模型(ASε-SVM).ASε-SVM为每个样本点赋一个单独的不敏感损失值,以此降低模型对包含较大噪声的样本点的依赖性,并从理论上证明了该方法可增强模型部分的泛化性能.将ASε-SVM与ε-SVM共同应用于处理一个数值算例和一个汽车销售预测实例中,仿真实验结果表明,ASε-SVM是有效可行的,可获得比ε-SVM更精确的预测结果.
Aiming at the feature of data including the noise of product sales timing, a support vector machine model (ASε-SVM) based on adaptive segment loss function is proposed. ASε-SVM assigns a separate insensitive loss value to each sample point, Reduce the dependence of the model on the sample points containing more noise, and theoretically prove that this method can enhance the generalization performance of the model part.Application of ASε-SVM and ε-SVM to deal with a numerical example and a car In the sales forecasting example, the simulation results show that ASε-SVM is effective and feasible, and more accurate prediction results than ε-SVM can be obtained.