Linguistic summaries of time series using fuzzy sets and their applications.

Presentation by Anna Wilbik, Thursday 23 May 14:30-15:30 in Paviljoen K.16.

Title: Linguistic summaries of time series using fuzzy sets and their applications.

Abstract

I present the concept of linguistic summaries of time series. Basically, the linguistic summaries are interpreted in terms of the number or proportion of elements possessing a certain property. I provide two applications of linguistic summaries: one for analysis of performance of mutual (investment) fund, second in eldercare domain.

In case of mutual funds, summaries of time series we propose refer in fact to the summaries of trends (segments) identified with straight line segments of a piecewise linear approximation of time series. Such summaries exemplified by “among all segments, most are short” or in a more sophisticated form by “among all long segments, most are slowly increasing” can be easily interpreted using Zadeh’s calculus of linguistically quantified propositions.

In a case study from the eldercare domain the goal is to compare different nighttime patterns for change detection. The reasons for studying linguistic summaries for eldercare are twofold: first, linguistic summaries are the natural communication tool for health care providers in a decision support system, and second, due to the extremely large volume of raw data, these summaries create compact features for an automated reasoning for detection and prediction of health changes as part of the decision support system. We have developed a metric distance for our particular form of linguistic summaries. This has allowed the creation of linguistic prototypes from clusters of summaries over some temporal range. Using that, we present a method for detecting anomalies, as observations considerably different from the linguistic prototypes in a moving temporal window. An example demonstrate the utility of this approach.