A supervised descriptive local pattern mining approach to the discovery of subgroups with exceptional survival behaviour

• Ano de defesa: 2021
• Autor(a) principal: MATTOS, Juliana Barcellos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Pernambuco; UFPE; Brasil; Programa de Pos Graduacao em Ciencia da Computacao
• Palavras-chave em Português: Inteligência computacional; Mineração de modelos
• Link de acesso: https://repositorio.ufpe.br/handle/123456789/44957

A variety of works in the literature strive to uncover the factors associated with survival behaviour. However, the computational tools to provide such information are global models designed to predict if or when a (survival) event will occur. When addressing the problem of explaining differences in survival behaviour, those approaches rely on (assumptions of) predictive features followed by risk stratification. In other words, they lack the ability to discover local exceptionalities in the data and provide new information on factors related to survival. In this work, we aim at providing a computational tool to identify the different (unusual) survival responses that may occur in a population of individuals and provide straightforward information about the circumstances related to such responses. We approach such a problem from the perspective of supervised descriptive pattern mining to discover local patterns associated with different survival behaviours. Hence, we introduce an Exceptional Model Mining (EMM) framework to provide straightforward characterisations of subgroups presenting unusual survival models, given by the Kaplan-Meier estimates. In contrast to the greedy search heuristics prevalent among EMM approaches, we employ stochastic optimisation and introduce the first approach in the literature to explore the Ant-Colony Optimisation (ACO) meta-heuristics for the subgroup search. Thus, we tackle the problem of subgroup redundancy to provide a set of exceptional subgroups that are diverse in their descriptions, coverages and survival models. We conducted experiments on fourteen real-world data sets to assess the performance of our approach. In the results, we show that the framework presented is capable of discovering representative patterns with accurate unusual models and straightforward representations. Moreover, the discovered subgroups potentially capture survival behaviours existent in the data. The approach successfully tackles the problem of subgroup redundancy, providing a set of diverse (unique) exceptional (survival) subgroups. Our framework outperforms the other existent approaches to provide characterisations over unusual survival behaviours regarding the descriptive aspect of its results and diversity of its findings.