Transportation mode classification through ordinal patterns with amplitude information

• Ano de defesa: 2020
• Autor(a) principal: Isadora Cardoso Pereira da Silva
• Orientador(a): Heitor Soares Ramos Filho
• Banca de defesa: Alejanndro César Frery Orgambide, Jefersson Alex dos Santos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Minas Gerais
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: ICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO
• País: Brasil
• Palavras-chave em Português: Transportation mode classification; Time Series Classification; Ordinal Patterns
• Link de acesso: http://hdl.handle.net/1843/33413; https://orcid.org/0000-0002-7681-7653

The infrastructure of cities is experiencing significant stress, since the demand for basic resources (such as transport, education, healthcare, etc) is outstripping supply. This is happening due to the disorderly growth caused by migration and increasing of the world’s population. Therefore, the scientific and industrial communities are investing in solutions based on human mobility that can provide a more sustainable development and reduce several commuting problems, such as traffic jam, in order to improve the life quality of humans. A critical step to achieve such goals is to characterize the transportation mode used. It is paramount the development of technologies that extract this kind of information, without the active participation of users on the act, hence avoiding inaccurate and incomplete data. In this context, this dissertation aim to develop a framework that, from user location data, can identify the transportation modes used. This framework contains four step: (i) segmentation; (ii) feature extraction; (iii) data transformation; and (iv) classification. More attention is given to the third step, where we propose a data transformation based on Ordinal Patterns (OP) probability distribution, capable of extracting the amplitude information presented in data – called Ordinal Pattern with Amplitude Information (OPAI). In our experiments, performed in real data, we show that OPAI presents superior classification results compared to OP transformation, a gain of about 10% of accuracy, indicating that OPAI is a technique with potential for the identification of transportation mode.