Coelho, José Romana Baptista, 1986-Tomás, Helena Isabel Aidos LopesGomes, Bernardo Policarpo2022-12-292022-12-2920222022http://hdl.handle.net/10451/55538Tese de mestrado, Engenharia Informática , 2022, Universidade de Lisboa, Faculdade de CiênciasWith the increase in the amount of vehicles and the population in big cities, problems related to traffic jams, traffic congestion and pollution arise with it. A lot of investigation has been done to try and solve or, at least, mitigate this problem. Governments are trying to mitigate traffic congestion and traffic jams by better understanding traffic, its characteristics and its patterns and getting insights about traffic. The purpose of this research is to create general traffic indicators for the city of Lisbon and, to do so, we will apply state of the art methods to a dataset of traffic from the city of Lisbon, provided by Camara Municipal de Lisboa ˆ that contain traffic data from the years of 2019 and 2020. We discuss the several types of data used in this type of problem, the pre-processing techniques used to transform the data, the several state of the art methods used for both prediction of traffic flow, and classification of different traffic situations, and also the performance metrics used to evaluate results. We make an exploratory and a more complex analysis to the provided data and also a discussion about the influence of the Covid-19 pandemic on the data and the problems that this could bring. We explain all the pre-processing and data cleaning techniques we used to handle the data, all the prediction models used, as in LSTM and ARIMA, and all the classification models used, as in Decision Tree Classifier and SVM. For the prediction task, LSTM obtained a mean RMSE of 10.493, while ARIMA got a mean RMSE of 38.722. For the classification task, DTC got a mean accuracy of 96.7%, while SVM got a mean accuracy of 88.6%.engtrânsitoprevisãoclassificaçãoLSTMDTCTeses de mestrado - 2022master thesis203203151