Linguistic Summarization of Time Series Data using Genetic Algorithms

Abstract

In this paper, the use of an evolutionary approach when obtaining linguistic summaries from time series data is proposed. We assume the availability of a hierarchical partition of the time dimension in the time series. The use of natural language allows the human users to understand the resulting summaries in an easy way. The number of possible final summaries and the different ways of measuring their quality has taken us to adopt the use of a multi objective evolutionary algorithm. We compare the results of the new approach with our previous greedy algorithms.

Linguistic Summarization of Processes – a research agenda

“…Last but not least, the common challenge for all types of linguistic summaries is the generation process (cf. Kacprzyk and Wilbik [15], Pilarski [26], Castillo-Ortega et al [9]). There are two main issues: efficiency and completeness.…”

On the role of linguistic descriptions of data in the building of natural language generation systems

“…The given example is made on data about patient inflow in medical centers, from which summaries such as "Most of the days with cold weather patient inflow is low or very low" or "Most of the days of June, patient inflow is medium" are obtained. This use case was also explored in [87] using a genetic algorithm approach instead of the standard heuristic algorithms used to generate linguistic descriptions. Another interesting research by Castillo et al addresses the problem of obtaining hierarchical segmentations of time series data and their application in linguistic descriptions [103] (see Fig.…”

“…[82,78]) and meta-heuristic (e.g. [87,88]) approaches can be used to address the linguistic description search process.…”

A Framework for Generating Explanations from Temporal Personal Health Data

“…Linguistic summarization of time-series data has been used for various applications, such as elderly care [40], physical activity tracking [33], driving simulation environments [13], deforestation analysis [11], human gait study [1], periodicity detection [28], time-series forecasting [22], and generation of longer temporal "narratives" from neonatal intensive care data via the use of a neonatal ontology [14]. Other work includes the use of genetic algorithms [7] to generate linguistic summaries from time-series, and those that place emphasis…”