• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.45.2018.tde-20230727-113344
Document
Author
Full name
Lucas Tavares Short Cabral
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2018
Supervisor
Title in Portuguese
Consenso Monte Carlo em modelos BART: priori, agregação e predição
Keywords in Portuguese
Algoritmos
Big Data
Inferência Estatística
Método De Monte Carlo
Abstract in Portuguese
Esse trabalho visa estudar em um contexto de Big Data o Bayesian Additive Regression Tree (BART) quando amostrado por um Consensus Monte Carlo (CMC). O BART é um modelo de regressão não-paramétrica que utiliza da soma várias árvores binárias com profundidade regulariza- das, via priori, para construir funções preditivas. Já o CMC é uma maneira de combinar amostras de Monte Carlo de diferentes computadores (ou de partições dos dados) e gerar uma aproximação da posteriori dos dados completos. Os objetivos do trabalho são estudar o comportamento das pri- oris nesse contexto em relação a capacidade de predição do modelo e verificar como se comporta seleção de variáveis do BART em um CMC. Para atingir o objetivo propõem-se uma nova maneira de agregar os resultados dos diferentes BART2019s no CMC usando a correlação de Pearson combinada a variância das predições realizadas por cada BART individual como peso. Os resultados mostram que o CMC do modelo BART é capaz de selecionar e predizer tão bem quanto o BART, com maior escalabilidade.
Title in English
Consensus Monte Carlo on BART models: prior, aggregation and predition
Abstract in English
This work study the Bayesian Additive Regression Tree (BART) when sampled by a Consensus Monte Carlo. BART is a non-parametric regression model that sum the outcome of many binary trees with regularized depth, via prior, to make predictions. On the other hand, CMC is a method to aggregate MCMC samples from different computers (or data shards) to generate an approxi- mation to the full data posterior. The objectives for this work are: study the impact of the priors choice in prediction of BART with CMC, and it2019s capability of variable selection. To accomplish the objective we suggest a new way to aggregate the consensus results from each BART in a CMC. This new aggregation method is based on Pearson2019s correlation and the variance of the predictions to generate the weights. The results shows that CMC-BART is able to select variable and make prediction as good as BART, but with more scalability.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2023-07-27
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.