| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.53 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Com o desenvolvimento das tecnologias de Machine Learning deu-se uma revolução nos
processos de tomada de decisão de diversas indústrias. Para aumentar a disponibilidade desta
tecnologia para um publico mais geral existe um grande interesse na automatização progressiva
de operações de Machine Learning, AutoML.
Nesta dissertação em primeiro lugar utiliza-se uma adaptação do PRISMA para fazer uma
revisão do estado da arte para os conceitos do ML e as suas ferramentas de automatização. Em
seguida é realizada uma pequena avaliação e comparação de algumas dessas ferramentas, com
base na resolução de problemas simples, de modo a conseguir perceber as vantagens e
desvantagens práticas de cada uma dessas soluções. Por último é abordado e explicado o
processo de design, implementação e avaliação de uma aplicação de AutoML para problemas
de classificação.
With the development of Machine Learning technologies, a revolution has taken place in the decision-making processes of various industries. In order to increase the availability of this technology to the general public, there is great interest in the progressive automation of Machine Learning operations, AutoML. This dissertation first uses an adaptation of PRISMA to review the state of the art in ML concepts and its automation tools. This is followed by a short evaluation and comparison of some of these tools, based on solving simple problems, to understand the practical advantages and disadvantages of each of these solutions. Finally, the process of designing, implementing and evaluating an AutoML application for classification problems is discussed and explained.
With the development of Machine Learning technologies, a revolution has taken place in the decision-making processes of various industries. In order to increase the availability of this technology to the general public, there is great interest in the progressive automation of Machine Learning operations, AutoML. This dissertation first uses an adaptation of PRISMA to review the state of the art in ML concepts and its automation tools. This is followed by a short evaluation and comparison of some of these tools, based on solving simple problems, to understand the practical advantages and disadvantages of each of these solutions. Finally, the process of designing, implementing and evaluating an AutoML application for classification problems is discussed and explained.
Description
Keywords
Machine learning Automation Machine learning pipeline AutoML Automatização