点击次数：

论文类型：期刊论文

发表时间：2019-04-01

发表刊物：INFORMATION SCIENCES

收录刊物：SCIE、SSCI、Scopus

卷号：478

页面范围：476-498

ISSN号：0020-0255

关键字：Granular computing; Granular fuzzy modeling; Linguistic models; Interval-valued time series

摘要：Modeling interval-valued time series (ITS) is an ongoing timely issue in the domain of time series analysis. Many researchers proposed diverse numeric models showing better performance of these models at the numeric level. However, a question how to establish a linguistic model of ITS exhibiting both high accuracy and interpretability is rarely studied. In this study, a linguistic modeling approach of ITS is presented by following the design methodology of granular computing. The crux of the proposed approach involves the formation of granular codebook consisting of a series of fundamental concepts describing amplitude characteristics of ITS, the granular expression mechanism of ITS and the realization of granular mapping based on multilayer perceptrons (MLPs). Further, the topology of neural network of the formed linguistic model is also presented to show the characteristics of layered processing information of granular computing (GrC). Experimental studies are reported for several publicly available financial ITS showing different dynamic characteristics, which offer a useful insight into the effectiveness of the proposed approach as well as reveal the impact of their parameter on the performance of the established linguistic model. (C) 2018 Elsevier Inc. All rights reserved.