Search

Publisher Copyright: © 2022. The Author(s). BACKGROUND: Tuberculosis (TB) diagnosis and treatment delays increase the period of infectiousness, making TB control difficult and increasing the fatality rates. This study aimed to determine the evolution of health care service delay (time between the patient's first contact with the health service and the diagnosis/start of treatment) and patient delay (time between onset symptoms date and the date of first contact with health services) for Pulmonary Tuberculosis (PTB) in Portugal between 2008 and 2017 across different regions, age groups and gender. METHODS: An exploratory analysis was performed, trends of both delays were studied, and 36 months forecasts were generated. We used the permutation test to test differences between groups and the Seasonal and Trend decomposition using Loess (STL) method and Autoregressive Integrated Moving Average (ARIMA) models for forecasting for both Health and Patient delays. We used data from notified PTB cases in mainland Portugal between 2008 and 2017, provided by the national surveillance system. RESULTS: Health delays remained relatively constant while patient delays increased. Females had significantly higher health delays in some regions. Individuals older than 64 had higher health delays than younger individuals, while patient delay for working-age individuals between 15 and 64 years old, presents higher patient delay. CONCLUSIONS: Forecasts presage that the upward trend of the delays is unlikely to fall in the coming years. It is important to understand the evolution of the delays and predict how these will evolve. Our understanding of the delays behaviours will contribute to better health policies and resources allocation. publishersversion published

Submitted by Maria da Luz Antunes (mluz.antunes@estesl.ipl.pt) on 2021-03-26T16:06:49Z No. of bitstreams: 1 Structured additive regression modeling of pulmonary tuberculosis infection.pdf: 806280 bytes, checksum: eb12663db17b1f4ea5b431381453408d (MD5) Made available in DSpace on 2021-03-26T16:06:50Z (GMT). No. of bitstreams: 1 Structured additive regression modeling of pulmonary tuberculosis infection.pdf: 806280 bytes, checksum: eb12663db17b1f4ea5b431381453408d (MD5) Previous issue date: 2020-02 info:eu-repo/semantics/publishedVersion

Submitted by azevedo@ufp.pt (azevedo@ufp.pt) on 2019-02-27T11:31:26Z No. of bitstreams: 1 PPG_28167.pdf: 47304833 bytes, checksum: 7a2b1ec774793990eb28bdd17968985e (MD5) Approved for entry into archive by halves@ufp.pt (halves@ufp.pt) on 2019-02-27T11:43:56Z (GMT) No. of bitstreams: 1 PPG_28167.pdf: 47304833 bytes, checksum: 7a2b1ec774793990eb28bdd17968985e (MD5) Made available in DSpace on 2019-02-27T11:43:56Z (GMT). No. of bitstreams: 1 PPG_28167.pdf: 47304833 bytes, checksum: 7a2b1ec774793990eb28bdd17968985e (MD5) Previous issue date: 2018-09-19

Data insufficiency hampered the academic research of discrimination based on sexual orientation. This is a particular concern in Portugal, a country that in spite of the strong legal recognition of homosexuals still scores low in their acceptance. Resorting to a selfdesigned web-survey, this study provides for the first time academic investigation of homosexuals in Portugal and contributes with an evaluation of wage discrimination in the primary employment. The empirical results point in the direction of absence of discrimination, but there is imprecise and small evidence that some homosexual individuals may be subject to discrimination.

Tese de mestrado, Medicina do Adolescente, Faculdade de Medicina, Universidade de Lisboa, 2015 Submitted by Lucília Mateus (luciliamateus@fm.ul.pt) on 2016-04-01T09:58:01Z No. of bitstreams: 1 10902_Tese.pdf.pdf: 952470 bytes, checksum: 4796a981b0d65a8c6d0644718a0a3e47 (MD5) Made available in DSpace on 2016-04-01T11:09:44Z (GMT). No. of bitstreams: 1 10902_Tese.pdf.pdf: 952470 bytes, checksum: 4796a981b0d65a8c6d0644718a0a3e47 (MD5) Previous issue date: 2015

A água é um bem essencial e escasso nos dias de hoje, pelo que é necessário conservá-lo e distribuí-lo de maneira eficiente e sustentável. Com a ajuda dos novos medidores digitais é possível medir os consumos com uma cadência horária, e através de técnicas de Machine Learning, produzir informações sobre padrões de consumo, comportamento dos consumidores, deteção de eventos nas redes de distribuição de água e feedback aos clientes. A deteção de anomalias, como fugas de água domésticas e na rede de saneamento, avaria dos medidores e fraude, são algumas das causas que provocam desperdícios de água e que muitas autarquias não têm como detetar de forma eficaz. A presente dissertação tem como objetivo desenvolver uma metodologia, baseada em séries temporais, para a previsão de consumos de água e a deteção de anomalias, recorrendo a dados dos consumidores domésticos, de uma zona situada na Quinta do Lago, Algarve. Para isso, são utilizados métodos de Clustering para agrupar os consumidores com padrões de consumo idênticos, e modelos de Machine Learning, como o ARIMA, Neural Networks e Random Forest, utilizados para prever os consumos de água. São apresentados os resultados de vários modelos para previsão de consumos e, por fim, são desenvolvidos dashboards que auxiliam na deteção de possíveis anomalias. Water is an essential and limited resource these days, so it is necessary to conserve and distribute it efficiently and sustainably. With the help of the new digital meters, it is possible to measure consumption on an hourly basis, and through Machine Learning techniques, produce information on consumption patterns, consumer behavior, detection of events in water distribution networks and feedback to customers. The detection of anomalies, such as domestic water leaks and in the sanitation network, meter failure and fraud, are some of the causes that cause water waste and that many municipalities are unable to detect effectively. This dissertation aims to develop a methodology, based on time series, for forecasting water consumption and detecting anomalies, using data from domestic consumers in an area located in Quinta do Lago, Algarve. For this, Clustering methods are used to group consumers with identical consumption patterns, and Machine Learning models such as ARIMA, Neural Networks and Random Forest, used to predict water consumption. The results of several models for forecasting consumption are presented, and finally, dashboards are developed to help detect possible anomalies.

Recentes avanços em campos como deep learning e computer vision, trouxeram novas possibilidades à área do retalho. Tais avanços forneceram novas ideias e capacidades às lojas de explorarem novos tipos de informação, devido às múltiplas câmaras implementadas no estabelecimento. Nesta dissertação, é desenvolvida uma investigação que estuda possíveis comportamentos de clientes que esses estabelecimentos podem explorar a partir da análise de imagens. Neste seguimento, é necessário perceber como se pode analisar e tirar proveito das mesmas e ter a capacidade de reconhecer intervenientes no processo de compra, através da parametrização do melhor modelo, utilizando técnicas de deep learning. Com os objetivos da investigação em mente, é realizada uma extensa pesquisa através de uma Revisão Sistemática da Literatura na área de Big Data aplicada a computer vision, deep learning e os métodos de análise de sensores. Adicionalmente, é efetuada uma análise aos resultados de um questionário, aplicado a consumidores nacionais, que pretenderia criar empatia com os mesmo, de modo a tentar traduzir as suas posições em propostas de valor. Paralelamente, foram recolhidos e analisados dados de imagem provenientes de um retalhista português e utilizados os algoritmos Yolo e Hidden Markov Model, considerando a sua rapidez e características, mas também a presença em estudos modernos. Os resultados permitiram identificar pessoas com uma accuracy média de 85%. Foram ainda construídos mapas de frequências e heat maps, com o principal contributo de enaltecer padrões de consumo e afluências na loja, para aferir o impacto que este tipo de informação pode ter na experiência do consumidor. Recent advances in fields such as deep learning and computer vision have brought new possibilities in the retail area. Such advances have provided new insights and capabilities for shops to exploit new types of information due to the multiple cameras deployed in the establishment. In this dissertation, an investigation is developed that studies possible customer behaviours that these establishments can exploit from image analysis. In this follow up, it is necessary to understand how one can analyse and take advantage of them and have the ability to recognise actors in the shopping process by parameterising the best model using deep learning techniques. With the research objectives in mind, an extensive literature review is carried out in the area of Big Data applied to computer vision and deep learning, but also in sensor analysis methods. Additionally, a survey was implemented to a portuguese consumers sample that intended to create empathy with consumers in order to understand their wishes to try to translate them into value propositions. In parallel, image data from a portuguese retailer was collected and the Yolo algorithm was used together with Hidden Markov Model to analyse it, considering its speed and characteristics, but also its presence in modern studies. The results allow identifying people with an average accuracy of 85%. Frequency maps and heat maps were also generated, with the main contribution of highlighting consumption patterns and affluence in the shop, to assess the impact that this type of information can have on consumer experience.