Previsão do índice de desenvolvimento humano e da expectativa de vida na América Latina por meio de técnicas de mineração de dados

• Ano de defesa: 2016
• Autor(a) principal: Santos, Celso Bilynkievycz dos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Ponta Grossa; Brasil; Programa de Pós-Graduação em Engenharia de Produção; UTFPR
• Palavras-chave em Português: Indicadores sociais; Desenvolvimento social; Qualidade de vida - América Latina; Mineração de dados (Computação); Social indicators; Progress; Quality of life - Latin America; Data mining; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO; Engenharia de Produção
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/2325

The predictability of quality of life indicators can contribute to the projection of dependent variables, help decision-making processes to support public policies and justify the scenario experienced by the countries and the world. Aim: This study aimed to predict the Human Development Index (HDI) and life expectancy (LE) in Latin American countries in the period of 2015–2020 using data mining techniques. Methodology: The study followed the steps of Knowledge Discovery in Database (KDD) processes. During the data mining KDD step, the performance of different algorithms with function-based learning paradigms was analyzed. From the algorithm with the best performance, 748 prediction models of univariate and two multivariate were developed to predict the HDI of 187 countries and their results were compared with the last reports from the United Nations Development Program (UNDP) in order to define the most efficient model. The results of these prediction tests were compared with 44 univariate Autoregressive Integrated Moving Average (ARIMA) models. From the definition of the best algorithm of data mining and model, the prediction of HDI and LE for Latin American countries from 2015 and 2020 was done. Results: The SMOReg algorithm and the multivariate models presented the best performance in the tests during the experiment. The average growth in HDI and LE predicted for Latin American countries tend to increase in the period analyzed, 4.99±3.90 % and 2.47±0.09 years, respectively. Conclusion: Multivariate experiences allow better learning of algorithms, increasing their prediction. Mining data techniques present better quality in the predictions compared to Autoregressive Integrated Moving Average (ARIMA), which is the most popular technique. The predictions suggest an average growth in HDI and LE in Latin American countries compared to the world average.