交互式Pareto前沿可视化决策

Interactive visualization decision-making method of Pareto front

针对现有高维多目标可视化技术不能有效显示决策偏好信息这一难题,提出一种基于径向轴图的交互式Pareto前沿可视化决策方法。进行标准化适应度函数处理,消除量纲差距;根据交互更新的轴向量矩阵建立最小二乘问题模型,通过Moore-Penrose广义逆矩阵方法求解投影点,从而获得Pareto前沿分布图。该方法能够有效显示目标各维度信息,直观呈现决策偏好,动态表现目标性能趋势。通过实验与传统可视化技术对比分析,验证了该方法的有效性和实用性。

Considering the inability of current visualization technology of high dimension multi-objective to effectively display preference information of decision-making,a new graphical visualization method of interactive n-dimension Pareto front method based on radial axes plots was proposed.Firstly,the fitness functions were standardized to eliminate dimensional gaps.Secondly,the least square problem model was established according to the interactively updated axes matrix and solved by the Moore-Penrose pseudoinverse method.Sequentially the diagram of dynamic Pareto front distribution was obtained,which effectively shows the data information of each dimension objective,the variation tendency towards performance and the preference information for decision-makers.Finally,the visualization model is devised and the extensive experimental analysis is provided to investigate the Pareto optimal solutions for decision-makers.