多特征不确定时间序列的关联趋势分析

针对不确定信息的相似性度量方法无法充分反映信息之间的关联情况,提出了直觉模糊集关联趋势分析法(RTIFS法)。利用直觉模糊集之间的距离表示不确定信息的差别,通过区间数与直觉模糊集之间的等价关系,利用区间数的距离计算直觉模糊集的关联度,最后应用集对分析法对序列间的关联趋势进行分类。RTIFS法将关联度计算的范围推广到不确定信息环境下,并给出多特征序列关联趋势的分类结果。实验结果表明,RTIFS法的分类准确率较高,算法运行时间短。

Analysis approach of relevance trend of time series withmultiple uncertain features

In view of the deficiency of means to directly calculate the relational degree between time series of uncertain information, this paper proposed the approach of the relevant trend analysis between series of intuitionistic fuzzy sets. The approach, firstly, quantified the difference between uncertain information by using the distance of intuitionistic fuzzy sets. Secondly, using equivalence relation between IFS and internal numbers, calculated the relevant degree of time series, on the basis of definition of distance between time series of interval numbers. Finally, classified the relevance trend of uncertain information series through the method of set pair analysis. The approach expanded the application range of relational degree from crisp numbers to uncertain environment expressed by intuitionistic fuzzy sets, and it classified the degree of relevance trend between time series with multiple uncertain features. Compared with the C-means algorithm and the simulation anneal algorithm, experiment results indicate, by using this method, the accurate rate of the algorithm is higher; and the false alarm rate and the missing alarm rate are both lower; furthermore, it reduces the running time effectively.