Um dos objetivos mais importantes da análise de dados foca-se cada vez mais na aplicação de análises estatísticas para tentar prever a reação da sociedade a um novo produto. A indústria cinematográfica é uma das maiores e mais importantes indústrias de entretenimento do mundo, gerando milhares de milhões de euros todos os anos. O objetivo deste projeto é criar um sistema que apoie a decisão na criação de um novo filme, usando técnicas de aprendizagem automática (machine learning) de forma a minimizar o risco do investimento. O sistema prevê o sucesso do filme baseado nas críticas recebidas pelo mesmo, tendo sido usados datasets fornecidos pelo IMDb para obter características antes e após lançamento dos filmes, para o treino e o teste dos algoritmos. Neste projeto o algoritmo de machine learning que obteve a melhor previsão foi o Gradient Boosting Regressor tendo apresentado uma precisão de aproximadamente 88%.

The characterisation of the seismic vulnerability of historical constructions represents a complex problem in which the typological variability, the difficulty of performing reliable large-scale assessments and dealing with a large database all play a role. Nevertheless, reducing the uncertainty regarding the structural vulnerability of the existing building stock (mostly for small and/or isolated human settlements) is key for risk assessment and management. The present work proposes a novel approach based on the integration of a series of open-source tools for assembling a vulnerability-oriented database that is linked to a series of external services for increasing its capabilities. The database was implemented in a Geographical Information System (GIS) environment and contains the survey of a seismic vulnerability index for masonry constructions based on an adapted version of the GNDT-II approach. A customised Python-based software for reading, managing and editing the database is herein presented. This program allows the execution of the most typical operations with no assistance from the GIS environment, facilitating user interaction. Furthermore, the calculations regarding the vulnerability index and levels of damage have been implemented in this program. Alternatives for distributing the database are implemented and discussed, such as cloud-based distribution and the use of the Transactional Web Feature Service (WFS-T) protocol for its virtual publishing. The entire framework herein presented is a replicable and feasible workflow that can be set even with reduced infrastructure, allowing a progressive enlargement.