Datamining

WORKING PAPER SERIES

CEEAplA WP No. 10/2009

Extrair Conhecimento de Bases de Dados: O caso dos Provérbios
Armando B. Mendes Günther Funk Gabriela Funk

August 2009

Universidade dos Açores Universidade da Madeira

Extrair Conhecimento de Bases de Dados: O caso dos Provérbios

Armando B. Mendes
Universidade dos Açores (DM) e CEEAplA

Günther Funk
Universidade dos Açores (DM) e IELT

Gabriela Funk
Universidade dos Açores (DLLM) e IELT

Working Paper n.º 10/2009 Agosto de 2009

CEEAplA Working Paper n.º 10/2009 Agosto de 2009

RESUMO/ABSTRACT
Extrair Conhecimento de Bases de Dados: O caso dos Provérbios

For data management activities in a project for proverbial sentences identification, a data base has being assembled during several years. This data base collects, in the moment of this study, information about 25.000 idiomatic sentences, including more than one thousand valid answers for proverbial sentences recognition surveys. In this article a project is described with the purpose to extract knowledge from this data base, in order to better characterize the individuals participating in the surveys about their level of proverbial recognition and the influence of the locations they have been living. In order to reach the study objectives we use data mining methodologies including: data preparation and preprocessing, data cleansing, and data reduction techniques. This data preparation stage is carefully described because we believe this is sometimes forgotten in statistical data mining studies and is a fundamental step to attain any data mining study objective. For data analysis, after a denormalized file is produced, we use linear regression models and regression trees with two different algorithms. The descriptive results are compared with paremiology domain knowledge, with some unexpected conclusions. Keywords: knowledge generation; data mining; proverbs; data preparation and pre-processing; regression trees.

Armando B. Mendes