Parametric, semiparametric regressions, and machine learning algorithms

• Grant number: 24/10798-4
• Support Opportunities: Scholarships abroad - Research
• Start date: March 01, 2025
• End date: February 28, 2026
• Field of knowledge: Agronomical Sciences - Agronomy
• Principal Investigator: Edwin Moises Marcos Ortega
• Grantee: Edwin Moises Marcos Ortega
• Host Investigator: Victor Hugo Lachos Davila
• Host Institution: Escola Superior de Agricultura Luiz de Queiroz (ESALQ). Universidade de São Paulo (USP). Piracicaba , SP, Brazil
• Institution abroad: University of Connecticut (UCONN), United States

Abstract

Currently, large amounts of data are generated from different areas of knowledge, with a variety of characteristics. In this sense, alternative methods are necessary to adequately and efficiently accommodate such particularities and to obtain as much information as possible regarding the data of interest. The aim of this project is to describe different regression models considering the generalized odd log-logistic-G family. These regression models are intended to accommodate characteristics such as heterogeneity of variances, nonlinear relationships and different risk functions. The proposal aims to compare the predictive performance of models with machine learning algorithms: decision trees, random forests, random survival forests and XGBoost. Structural properties of the new distributions will be provided. Applications will be carried out in different areas of research (agronomy, engineering, biology and chemistry, among others), illustrating the usefulness of the new models. As for predictive capacity, the performance of these distributions will be compared to machine learning algorithms through simulation studies and applications to real data.