abstract: Driving is the coordinated operation of mind and body for movement of a vehicle, such as a car, or a bus. Driving, being considered an everyday activity for many people, still has an issue of safety. Driver distraction is becoming a critical safety problem. Speed, drunk driving as well as distracted driving are the three leading factors in the fatal car crashes. Distraction, which is defined as an excessive workload and limited attention, is the main paradigm that guides this research area. Driver behavior analysis can be used to address the distraction problem and provide an intelligent adaptive agent to work closely with the driver, fay beyond traditional algorithmic computational models. A variety of machine learning approaches has been proposed to estimate or predict drivers’ fatigue level using car data, driver status or a combination of them. Three important features of intelligence and cognition are perception, attention and sensory memory. In this thesis, I focused on memory and attention as essential parts of highly intelligent systems. Without memory, systems will only show limited intelligence since their response would be exclusively based on spontaneous decision without considering the effect of previous events. I proposed a memory-based sequence to predict the driver behavior and distraction level using neural network. The work started with a large-scale experiment to collect data and make an artificial intelligence-friendly dataset. After that, the data was used to train a deep neural network to estimate the driver behavior. With a focus on memory by using Long Short Term Memory (LSTM) network to increase the level of intelligence in two dimensions: Forgiveness of minor glitches, and accumulation of anomalous behavior., I reduced the model error and computational expense by adding attention mechanism on the top of LSTM models. This system can be generalized to build and train highly intelligent agents in other domains. / Dissertation/Thesis / Doctoral Dissertation Computer Engineering 2020

Electronic health record (EHR) systems are used by medical providers to streamline the workflow and enable sharing of patient data with different providers. Beyond that primary purpose, EHR data have been used in healthcare research for exploratory and predictive analytics. EHR data are heterogeneous collections of both structured and unstructured information. In order to store data in a structured way, several ontologies have been developed to describe diagnoses and treatments. On the other hand, the unstructured clinical notes contain various more nuanced information about patients. The multidimensionality and complexity of EHR data pose many unique challenges and problems for both data mining and medical communities. In this thesis, we address several important issues and develop novel deep learning approaches in order to extract insightful knowledge from these data. Representing words as low dimensional vectors is very useful in many natural language processing tasks. This idea has been extended to medical domain where medical codes listed in medical claims are represented as vectors to facilitate exploratory analysis and predictive modeling. However, depending on a type of a medical provider, medical claims can use medical codes from different ontologies or from a combination of ontologies, which complicates learning of the representations. To be able to properly utilize such multi-source medical claim data, we propose an approach that represents medical codes from different ontologies in the same vector space. The new approach was evaluated on the code cross-reference problem, which aims at identifying similar codes across different ontologies. In our experiments, we show the proposed approach provide superior cross-referencing when compared to several existing approaches. Furthermore, considering EHR data also contain unstructured clinical notes, we also propose a method that jointly learns medical concept and word representations. The jointly learned representations of medical codes and words can be used to extract phenotypes of different diseases. Various deep learning models have recently been applied to predictive modeling of Electronic Health Records (EHR). In EHR data, each patient is represented as a sequence of temporally ordered irregularly sampled visits to health providers, where each visit is recorded as an unordered set of medical codes specifying patient's diagnosis and treatment provided during the visit. We propose a novel interpretable deep learning model, called Timeline. The main novelty of Timeline is that it has a mechanism that learns time decay factors for every medical code. We evaluated Timeline on two large-scale real world data sets. The specific task was to predict what is the primary diagnosis category for the next hospital visit given previous visits. Our results show that Timeline has higher accuracy than the state of the art deep learning models based on RNN. Clinical notes contain detailed information about health status of patients for each of their encounters with a health system. Developing effective models to automatically assign medical codes to clinical notes has been a long-standing active research area. Considering the large amount of online disease knowledge sources, which contain detailed information about signs and symptoms of different diseases, their risk factors, and epidemiology, we consider Wikipedia as an external knowledge source and propose Knowledge Source Integration (KSI), a novel end-to-end code assignment framework, which can integrate external knowledge during training of any baseline deep learning model. To evaluate KSI, we experimented with automatic assignment of ICD-9 diagnosis codes to clinical notes, aided by Wikipedia documents corresponding to the ICD-9 codes. The results show that KSI consistently improves the baseline models and that it is particularly successful in rare codes prediction. / Computer and Information Science

Made available in DSpace on 2011-05-04T12:36:19Z (GMT). No. of bitstreams: 0 Previous issue date: 2010 / Trata-se de um estudo de caso que examina a prática do acolhimento em uma unidade básica de saúde da cidade de Juiz de Fora MG, numa ótica qualitativa. O objeto de estudo deste trabalho é o acolhimento propriamente dito, numa exploração de suas relações com a Política Nacional de Humanização, com a mudança de modelo assistencial e, em última análise, no reflexo desta prática nos serviços que chamamos serviços de ponta, procurando entender como o acolhimento é traduzido / reinterpretado nas práticas cotidianas. O relato de experiência busca elementos de análise a partir da conversa com outros profissionais que participaram desse processo, através de entrevistas individuais. Partimos de três hipóteses elementares: 1) a Política Nacional de Humanização tem sofrido interpretações e traduções pelos profissionais de saúde do nível local; 2) as distorções na organização do sistema podem gerar conseqüências impremeditadas na medida em que contribuem para a reprodução de práticas ineficazes e embotadas no acolhimento, por falta de estímulos à mudança continuidade nas suas ações e 3) existem diferenças importantes entre as concepções do acolhimento em proposições teóricas de três diferentes propostas desenvolvidas no âmbito da política de saúde brasileira no período mais recente: o Projeto em Defesa da Vida (PDV), os programas de Humanização e a Política Nacional de Humanização (PNH) e o que é realizado na prática em nome do acolhimento. Concluímos que o acolhimento realizado na UBS Abrigo apresenta conseqüências impremeditadas que nem sempre condizem com a boa prática em saúde, que ele acontece impulsionado principalmente pelos profissionais de saúde ligados à residência multiprofissional, e que essa prática possui potencialidades para mudanças nos processos de trabalho e no modelo assistencial, onde muitas mudanças já se tornam reais, mas ainda evidencia alguns desafios a superar como a não-adesão de todos os profissionais na prática do acolhimento, a falta de comprometimento da gestão e as dificuldades relacionais. / This study regards the case which examines the practice of acceptance at a basic healthcare unit in Juiz de Fora City – Minas Gerais State, upon a qualitative view. The study object for this work is the acceptance as mentioned previously, exploring its relations to the National Politics of Humanization with changes in assisting model and, ultimately on the reflex of this practice on services denominated state-of –the-art services as we try to understand how acceptance is translated/ reinterpreted on daily practices. The report of experience seeks analysis elements as of conversation with other professionals who have taken part on such process by individual interviews. Three elementary hypothesis are the starting point: 1) the National Politics of Humanization has been through interpretations and translations by the healthcare professionals of a local level; 2) distortions in the system organization might generate unplanned consequences in so far as to contribute to the reproduction of ineffective and blunt practices in acceptance due to the lack of stimulation in continual change in its actions and 3) there are important differences among the conceptions of acceptance in theoretical propositions of three different proposals developed in the scope of Brazilian healthcare politics in a latter period: Life Protection Project (LPP), Humanization programs and the National Politics of Humanization (NPH) and what is actually performed for the sake of acceptance. It has been concluded that the acceptance performed in “Abrigo”`s BHU presents unplanned consequences which are not usually in accordance to the good health practice, which happens mainly by healthcare professionals related to multiprofessional residency, practice which has changing potentials in the working process and assisting model, where many changes have become a reality but still shows challenges to be overcome such as the non-adherence of all professionals to the acceptance practice, managers’ lack of commitment and relational difficulties.

Transformer models, delivering big improvement in AI text-models (NLP), are now being applied in Knowledge Tracing to track the knowledge of students over time. One of the first, SAINT, showed quite some improvement over the then SOTA results on the public EdNet dataset and caused an increase in research based on transformer-based models. In this paper, we firstly aim to reproduce the SAINT results on the EdNet dataset but are unable to report a similar performance as the original paper. This might be due to implementation details, which we were not able to completely reconstruct. We hope to pave the road for further reproducibility, as an increasingly important part of AI research. Furthermore, we apply the model to a company dataset much larger than any public dataset (more interactions, more exercises and more skills). Such a dataset is on the one hand more challenging (more skills mixed), and on the other hand, provides much more data (which should help our models). We compare the SAINT model and the seminal IRT model, and find that the SAINT model performance is 4% better in AUC but 1.7% worse in RMSE. Our experiments on window size suggest that transformer models still struggle with modelling beyond recent performance, and do not yet deliver the step-change observed in NLP. / Transformermodeller, som ger stora förbättringar av AI-textmodeller (NLP), används nu i Knowledge Tracing för att spåra elevernas kunskaper över tid. En av de första, SAINT, visade en hel del förbättring jämfört med de dåvarande SOTA-resultaten på den offentliga EdNet-datauppsättningen och orsakade en ökning av forskning baserad på transformerbaserade modeller. I denna artikeln siktar vi först efter att återskapa SAINT-resultaten på EdNet-datauppsättningen, men vi kan inte rapportera liknande prestanda som den ursprungliga uppsatsen. Detta kan bero på implementeringsdetaljer som vi inte kunde rekonstruera helt. Vi hoppas kunna bana väg för ytterligare reproduktioner, som en allt viktigare del av AI-forskningen. Dessutom tillämpar vi modellen på en företagsdatauppsättning som är mycket större än någon offentlig datauppsättning (fler interaktioner, fler övningar och fler färdigheter). En sådan datauppsättning är å ena sidan mer utmanande (mer blandad kompetens), men å andra sidan ger den mycket mer data (vilket borde hjälpa våra modeller). Vi jämför SAINT-modellen och den framträdande IRT-modellen och finner att SAINT-modellens prestanda är 4% bättre i AUC men 1,7% sämre i RMSE. Våra experiment på fönsterstorlek tyder på att transformermodeller fortfarande kämpar med modellering utöver de senaste prestanda och ännu inte levererar den stegförändring som observerats i NLP.