基于投影寻踪的时间序列多属性决策研究

On Multi-attribute Decision Making of Multidimensional Time Series Based on Projection Pursuit

针对属性权重完全未知且属性值为多维时间序列的评价决策问题,提出一种基于加速遗传算法-投影寻踪和多属性决策的混杂评价决策模型方法。该方法将首先利用投影寻踪方法对多维时间序列数据按照属性进行降维处理,以解决数据处理过程中"维数灾难"带来的影响,并使用加速遗传算法确定最佳投影方向作为属性权重;对于得到的具有时间序列特性的决策矩阵,使用指数平滑方法确定时间序列的系数权重,最后对备选方案进行综合评价决策。通过实例验证,证明该方法简捷实用,极大提高了工程决策效率和准确性。

A hybrid projection pursuit and multi-attribute decision making model method is presented to deal with the problem with the completely unknown attribute weights and the given attribute values in the time series data, base on the accelerating genetic algorithm. This projection pursuit method is used to process the dimensionality reduction, and the accelerating genetic algorithm is used to deter- mine the best projection direction. Then the exponential smoothing method is used to determine the weight coefficients of time series and make a comprhensive evaluation of the decision-making for alternative. The case analyses show that method is simple and practical, which greatly improve the efficiency and accuracy of engineering decision-making.