Tomás, Helena Isabel Aidos LopesFreitas, João Capucho Correia de2023-09-152023-09-1520232022http://hdl.handle.net/10451/59336Tese de Mestrado, Ciência de Dados, 2022, Universidade de Lisboa, Faculdade de CiênciasMost vehicles on the road today include some form of travel assistance through an insurance company. Travel assistance companies work 24/7 at the behest of their clients to ensure that, in case of breakdown or accident, their vehicles can be safely repaired, on the road or in a garage. To ensure adequate service levels and quick response times, companies in the travel assistance domain need to have concrete historical knowledge on a multitude of indicators, including the number of assistance requests per/day, their hourly distribution, the types of breakdowns, etc. All these factors constitute the basis on which day­to­day planning is thought out, from the number of tow­trucks required, the number of contact­centre operators available their schedule, information systems capacity, etc. In this thesis real data from a company in the assistance domain is processed, analysed and prepared to feed into several forecasting models, from the most basic ‘moving average’ model, to complex deep learning model to both evaluate their comparative requirements and overall performance. Models, from the least to the most complex are examined in the literature review section as well as particular aspects of time­series data and what do those aspects entail in a forecasting exercise. The goal is to understand how well the ‘novel’ machine learning models such as those based on recurrent neural networks compare to more established statistical approaches such as SARIMA (Seasonal, Auto­Regressive, Integrated, Moving Average). The knowledge gained from this exercise will serve as the foundation for future works in time­series prediction.engSérie­TemporalAprendizagem AutomáticaAssistência em ViagemDeep LearningTeses de mestrado - 2023master thesis203524365