Forecasting the human development index of 2013 and 2014 by means of data mining techniques in univariate and multivariate temporary series

• Autor(a) principal: Dos Santos, Celso Bilynkievycz
• Data de Publicação: 2019
• Outros Autores: Pedroso, Bruno, Guimarães, Alaine Margarete [UNESP], Pilatti, Luiz Alberto, Kovaleski, João Luiz
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/198244

The Human Development Index (HDI) is an indicator adopted by the World Health Organization to assess the quality of life of a given region. Its prediction can aid in planning and decision-making for policy guidance and advocacy to improve its development. This study predicted the HDI of 2013 and 2014 from forecasting data mining techniques in time series, completing all stages of the knowledge discovery process in databases. In the study, the predictive capacity of 376 models, two generic and 374 country specific, were evaluated. For the development of the models we used the SMOReg algorithm, executed in a Forecast programming interface application of the WEKA environment. The generic model was trained and tested with multivariate time series corresponding to the HDI records of 187 countries, while the specific models were developed from univariate time series corresponding to the individual historical behavior of the index in each country. The time variables used corresponded to historical and intermittent periods from 1980 to 2013 published in the report of the United Nations Development Program on 07/24/2014. In the empirical analysis it was verified that the multivariate models presented the best quality measures in the predictions. The predictions of the HDI 2013 were efficient, with no significant differences to published figures, while the predictions of HDI 2014 depend on comparison with figures released after the completion of the present study.