Parking guiding system with occupation prediction

• Autor(a) principal: Alface, Gonçalo Pereira
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/20239
• topic: Parking availability; Prediction; Mobile app; Probability; Parking management; Disponibilidade de estacionamento; Previsão; Aplicação móvel; Probabilidade; Gestão de estacionamento

Parking availability is an increasingly scarce and expensive resource within large cities, and this problem is considered to be one of the most critical transportation management system inside a big city. To approach this problem a proof of concept is presented as a way to guide a driver to the possible free parking lot through a prediction process using past data, correlated with traffic, weather conditions and time period features (year, month, day, holidays, and so on). A feature selection was performed by the study of data patterns, in order to understand the parking lot affluence and how certain features influence them, as well as to comprehend the sudden changes in the total occupation of the parking lot and which features really matter and have an impact on the total occupation. Those conclusions helped to create a robust and efficient predictive model in order to predict the parking lot availability rate more accurately. Three algorithms were used to build the predictive models as a way to test the most efficient and accurate one, namely Gradient Boosting Machine, Decision Random Forest and Neural Networks. Various types of models were tested with the aim of improving the results obtained, as well as understanding the impact of each of the processing of the data used. To complement this, a decision algorithm was created to guide the driver to the most optimal parking lot that presents better conditions, taking into account the location and driver characteristics, like the park more likely to have an available parking space, closer to the user’s current position or a more attractive price for the driver. Finally, these developments are integrated into a mobile application in order to work like an interface that the driver can interact.