考虑决策风险偏好的自适应支持向量机模型

在构造可以控制不平衡性的分类器,准确识别少数类,并使得决策者可以根据风险偏好与分类器进行交互,这对于人工智能在管理实践中的应用有极为重要的价值。提出了一种自适应支持向量机(ASVM)模型,使得类间隔最大化的同时,决策损失最小化,并基于粒子群优化算法(PSO)调节参数。该模型内在地考虑了数据不平衡性,并可为决策者与分类器的交互提供有效支持。实验及仿真结果表明,该模型在各种样本不平衡情况下都有很好性能,分类准确率显著地优于对比方法,而且相对稳定,并能很好地根据决策者的偏好控制分类器的决策风险。

Adaptive Support Vector Machines Considering Decision-making Risk Preference

This study aims to construct a classifier that can control the imbalance,identify minority class accurately,and make the decision maker be able to interact with the classifiers according to his own risk preference.It is of great value for the application of artificial intelligence in the management practice.This study proposed an adaptive support vector machine(ASVM) model,which minimized the decision-making cost while maximizing the margin between two classes,and adjusted parameters using the particle swarm optimization(PSO) approach.The model considered the imbalance of the datasets intrinsically,and provided effective support for decision maker interacting with the classifier.The experiment and simulation results demonstrate that the developed method has good performance in sample-imbalanced situation,its classification accuracy rate surpasses the comparative method evidently,and it is more stable than the latter.The developed method can control the decision-making cost very well.