A FUZZINNESS PARAMETER OPTIMIZATION METHOD TO EXTRACT THE OPTIMAL SET OF LINGUISTIC SUMMARIES FROM NUMERIC DATA

Abstract

Extracting a set of linguistic summaries from numeric data aims to produce summary sentences expressed in natural language that describe the hidden knowledge in the numeric dataset. A number of genetic algorithm models have been proposed to extract the optimal set of linguistic summaries, in which the algorithm model for extracting the set of linguistic summaries ensures the interpretability of the content of the summary sentences by applying genetic algorithm with greedy strategy gives quite good results. However, the determination of fuzziness parameter values of the algorithm model depends on the expert's intuition. In this paper, a method to optimize the fuzziness parameter values to improve the quality of the set of linguistic summaries extracted from numeric data is proposed. Experimental results with the creep database show that with the optimized fuzziness parameter values, the quality of the extracted set of linguistic summaries is better on three measures: fitness function value, average truth value and number of sentences with linguistic quantifier greater than a half.