Summarizing time series with probabilistic fuzzy quantifiers

In this work, we show that the probabilistic fuzzy quantification models fulfill a number of properties that make them appropriate for the summarization problem. We focus our investigation in the use of the FA model for data summarization. This model has a very solid theoretical behavior, since it is a (non-standard) Determiner Fuzzification Scheme. Based on the properties of the probabilistic models we propose a summarization algorithm based on linguistic and data mining criteria. Examples of use are provided in the domain of meteorology using daily temperature data sets, obtaining a plausible behavior of the summaries compatible with human expectations.

keywords: linguistic summaries, semi-fuzzy quantifiers