Predicting space occupancy for street paid parking

• Autor(a) principal: Silva, Marco Hélio Ramos
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/15415
• topic: Redes neuronais; Árvore de decisão; Métodos de previsão; Análise de dados; Área de estacionamento; Parking space forecasting; Street parking; Traffic; Parking guidance; Machine learning; Decision trees; Neural networks; Previsão de lugares de estacionamento; Estacionamento em rua; Gestão de tráfego; Assistência ao estacionamento; Aprendizagem automática

This dissertation discusses how to develop a prediction method for on-street parking space availability, using only historical occupancy data collected from on-street multi-space parking meters. It is analyzed how to transform the raw data into a dataset representing the occupancy and how can this information be used to detect when the parking spaces on a street are Vacant or Full. Attributes like weather conditions and holidays are added to the data, giving them more context and comprehension. After the data preparation and analysis, a prediction model is developed using machinelearning techniques that can forecast the availability of the parking spaces on a street at a specific day and on a given moment. For that, a classification method is implemented based on decision trees and neural networks, comparing both methods regarding results and development time. Particular attention is given to the algorithm parameters, to achieve the right balance between accuracy and computational time. The developed model proved effective, correctly capturing the different behavior of each street through the different weeks, and returning results useful to drivers searching for parking and to the business owners while monitoring their parking investments and returns.