Detalhes bibliográficos
Ano de defesa: |
2023 |
Autor(a) principal: |
Braz, Rafael dos Santos |
Orientador(a): |
Não Informado pela instituição |
Banca de defesa: |
Não Informado pela instituição |
Tipo de documento: |
Dissertação
|
Tipo de acesso: |
Acesso aberto |
Idioma: |
eng |
Instituição de defesa: |
Biblioteca Digitais de Teses e Dissertações da USP
|
Programa de Pós-Graduação: |
Não Informado pela instituição
|
Departamento: |
Não Informado pela instituição
|
País: |
Não Informado pela instituição
|
Palavras-chave em Português: |
|
Link de acesso: |
https://www.teses.usp.br/teses/disponiveis/55/55134/tde-10012024-173035/
|
Resumo: |
Models that can represent the behavior of systems, such as a Finite State Machine (FSM), are crucial for software development and maintenance as they serve as a base for several automated activities like testing, verification, validation, and refinement of systems. Contrasting their importance and value, models are usually complex and costly to obtain. One option to mitigate this problem is model inference which provides the possibility to automatically, or at least with little human interaction, learn a model that represents the behavior of a system. This process can be mainly classified into passive inference (builds models from examples of the behavior of a system) and active inference (builds models from interacting with the system). In this dissertation, we propose a method for learning separating sequences from traces (examples of a previously observed behavior of the system) and applying it to improve the process of model inference. A separating sequence is an input sequence capable of distinguishing a pair of distinct states of a machine by yielding different output sequences for each state. When a set of separating sequences distinguishes all pairs of distinct states in the FSM, it is called a characterization set, or W-set. Our proposed method receives a set of traces, processes them to extract all their k-length subsequences, and uses them to build a data structure called W-tree that summarizes the relevant observations of the systems behavior indicated in the traces. The methods output is a set of the n-best separating sequences that a model inference algorithm applies to improve its W-set and its inference process. We implemented our proposed method, integrated it with an active inference algorithm called hW -inference, and performed a case study in which we used 40 different traces. We observed that the proposed method could improve the learning process by 24%, on average, and up to 48% in the best-case setting. |