Most story-based games today have manually-scripted non-player characters (NPCs) and the scripts are usually simple and repetitive since it is time-consuming for game developers to script each character individually. ScriptEase, a publicly-available author-oriented developer tool, attempts to solve this problem by generating script code from high-level design patterns, for BioWare Corp.'s role-playing game Neverwinter Nights. The ALeRT algorithm uses reinforcement learning (RL) to automatically generate NPC behaviours that change over time as the NPCs learn from the successes or failures of their own actions. This thesis aims to provide a new learning mechanism to game agents so they are capable of adapting to new behaviours based on the actions of other agents. The new on-line RL algorithm, ALeRT-AM, which includes an agent-modeling mechanism, is applied in a series of combat experiments in Neverwinter Nights and integrated into ScriptEase to produce adaptive behaviour patterns for NPCs.

The Traffic alert and Collision Avoidance System (TCAS) is an aircraft collision avoidance system designed to prevent mid-air collisions. During an advisory, danger is imminent, and TCAS is assumed to have better, more up-to-date information than the ground operated air traffic control (ATC) facility. Following a TCAS RA is generally the safe course of action during an advisory. However, pilot compliance with RAs is surprisingly low. Results from a TCAS monitoring study show pilots are not complying with many TCAS advisories. As revealed by pilot-submitted Aviation Safety Reporting System (ASRS) reports, this noncompliance could be attributed, in part, to pilot confusion to TCAS operation as well as misunderstandings of the appropriate response to a TCAS issued advisory. This thesis details the development and evaluation of a TCAS training program intended to improve pilots' understanding of TCAS use for collision avoidance in a range of traffic situations. The training program integrated Demonstration Based and Event Based Training techniques. Its efficacy was analyzed in an integrated ATC-cockpit simulator study in which eighteen commercial airline pilots were asked to complete the TCAS training program and afterwards experienced twelve experimental traffic events. The trained pilots' performance was compared to the performance of 16 baseline pilots who did not receive the modified training. Overall, the training program did have a significant impact on the pilots' behavior and response to TCAS advisories. The measure Time Pilots First Achieved Compliance decreased with the trained pilots, as did the measure Autopilot Disconnect Time After RA Initiation. Trained pilots exhibited less aggressive performance in response to a TCAS RA (including a decrease in the measures Altitude Deviation Over Duration Of RA, Average Vertical Rate Difference, Maximum Vertical Rate Difference, and Maximum Vertical Rate). The measure Percent Compliance did not significantly vary between trained and baseline pilots, although trained pilots had a more consistent response in the traffic event with conflicting ATC guidance. Finally, on the post-experiment questionnaires, pilots commented on their increase in understanding of TCAS as well as an increase in their trust in the advisory system. Results of this research inform TCAS training objectives provided by the FAA as well as the design of TCAS training. Additionally, conclusions extend more broadly to improved training techniques for other similar complex, time-critical situations.

This Open Platform for Limit Protection guides the open design of maneuver limit protection systems in general, and manned, rotorcraft, aerospace applications in particular. The platform uses three stages of limit protection modules: limit cue creation, limit cue arbitration, and control system interface. A common set of limit cue modules provides commands that can include constraints, alerts, transfer functions, and friction. An arbitration module selects the best limit protection cues and distributes them to the most appropriate control path interface. This platform adopts a holistic approach to limit protection whereby it considers all potential interface points, including the pilots visual, aural, and tactile displays; and automatic command restraint shaping for autonomous limit protection. For each functional module, this thesis guides the control system designer through the design choices and information interfaces among the modules. Limit cue module design choices include type of prediction, prediction mechanism, method of critical control calculation, and type of limit cue. Special consideration is given to the nature of the limit, particularly the level of knowledge about it, and the ramifications for limit protection design, especially with respect to intelligent control methods such as fuzzy inference systems and neural networks. The Open Platform for Limit Protection reduces the effort required for initial limit protection design by defining a practical structure that still allows considerable design freedom. The platform reduces lifecycle effort through its open engineering systems approach of decoupled, modular design and standardized information interfaces. Using the Open Platform for Limit Protection, a carefree maneuver system is designed that addresses: main rotor blade stall as a steady-state limit; hub moment as a transient structural limit; and pilot induced oscillation as a controllability limit. The limit cue modules in this system make use of static neural networks, adaptive neural networks, and fuzzy inference systems to predict these limits. Visual (heads up display) and tactile (force-feedback) limit cues are employed. The carefree maneuver system is demonstrated in manned simulation using a General Helicopter (GENHEL) math model of the UH-60 Black Hawk, a projected, 53 degree field of view for the pilot, and a two-axis, active sidestick for cyclic control.

Η ανάπτυξη ενός ενσωματωμένου στο όχημα συστήματος υποβοήθησης του οδηγού για αποφυγή συγκρούσεων με άλλα οχήματα, βρίσκεται τελευταία στο επίκεντρο του ενδιαφέροντος. Στα συστήματα αυτά η αξιοπιστία αποτελεί ένα πολύ σημαντικό παράγοντα. Στην παρούσα εργασία αναπτύσσεται ένα σύστημα αναγνώρισης προπορευόμενου οχήματος βασισμένο σε εικόνες οι οποίες λαμβάνονται από βιντεοκάμερα που έχει ενσωματωθεί στο όχημα. Η μεθοδολογία την οποία επιλέξαμε να εργαστούμε περιλαμβάνει τον εντοπισμό των κόκκινων εικονοστοιχείων στην εικόνα και τη δημιουργία της αντίστοιχης δυαδικής εικόνας. Στη συνέχεια, με μορφολογική επεξεργασία της δυαδικής εικόνας εντοπίζουμε τις περιοχές που αντιστοιχούν στα πιθανά φανάρια του οχήματος. Με βάση τα σημεία των πιθανών φαναριών καθορίζουμε την περιοχή στην οποία περικλείεται το όχημα. Για την επιβεβαίωση της ύπαρξης οχήματος στην περιοχή αυτή, εκτελούμε έναν έλεγχο συμμετρίας βασιζόμενοι στην ομοιότητα των υποεικόνων και συνεχίζουμε με τον προσεγγιστικό υπολογισμό της απόστασής του. Τέλος, παρουσιάζουμε τα αποτελέσματα της μεθόδου, τα συμπεράσματα που προέκυψαν και προτείνουμε κατευθύνσεις για μελλοντικές βελτιώσεις. / Developing on-board automotive driver assistance systems aiming to alert drivers about possible collision with other vehicles has attracted a lot of attention lately. In these systems, robust and reliable vehicle detection is a critical step. In this work a vehicle detection system is developed based on video frames grabbed by a camera mounted on the vehicle. Vehicle detection is mainly based on the detection of its red rear-lights. First we detect all red pixels of the frame and create the corresponding binary image (mask). Then we detect the areas that possibly constitute vehicle’s rear-lights by performing morphological binary image processing. Based on that, we determine the boundary of the vehicle. To verify the presence of the vehicle in this area, we perform a symmetry test based on sub-image similarity. Finally, we present some experimental results and give directions for future improvements.

No description available.

This thesis examines the 20th century settlement and migration history of the Kwakwaka'wakw people of Alert Bay, British Columbia. Through an examination of three key shifts in settlement and migratory patterns, it traces how Aboriginal space and movement has been reconfigured in response to changing social, economic, and cultural landscapes. Each of these three shifts—village relocations, the decline of involvement in the capitalist and traditional food economies, and growing urban migration—reveals how Kwakwaka'wakw settlements and notions of community have changed in recent times. These shifts also indicate how innovative forms of migration have developed in, around, and between aboriginal communities. In addition to documenting some of the most profound changes in Aboriginal demographics since the early catastrophic disease epidemics, this thesis is also interested in continuity and the role local culture plays in shaping settlement and migratory behaviour. Drawing on Michel De Certeau's notion of "combinatory operations," I suggest that Aboriginal people have interpreted and responded to different types of displacement through operational systems shaped by contemporary reproductions of socio-cultural traditions. The thesis argues that the people of this community have responded to displacement with behaviour reflective of both innovation and cultural continuity. Until now, most research on aboriginal people has been either community- or urban-based. However, this focus on the terminal "beginning" or "end" of migration has tended to overshadow the role migration itself has played within Aboriginal society and culture. Rather than a process of suspension occurring between two points of settlement, migration itself is a socio-cultural phenomenon, itself no less important than the settlements upon which the process is anchored and defined.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Intrusion Detection Systems (IDS) are designed to monitor the computer network infrastructure against possible attacks by generating security alerts. With the increase of components connected to computer networks, traditional IDS are not capable of effectively detecting malicious attacks. This occurs either by the distributed amount of data that traverses the network or the complexity of the attacks launched against the network. Therefore, the design of Internet Early Warning Systems (IEWS) enables the early detection of threats in the network, possibly avoiding eventual damages to the network resources. The IEWS works as a sink that collects alerts from different sources (for example, from different IDS), centralizing and correlating information in order to provide a holistic view of the network. This way, the current dissertation describes an IEWS architecture for correlating alerts from (geographically) spread out IDS using the Case-Based Reasoning (CBR) technique together with IP Georeferencing. The results obtained during experiments, which were executed over the implementation of the developed technique, showed the viability of the technique in reducing false-positives. This demonstrates the applicability of the proposal as the basis for developing advanced techniques inside the extended IEWS architecture. / Sistemas de Detecção de Instrução (Intrusion Detection Systems IDS) são projetados para monitorar possíveis ataques à infraestruturas da rede através da geração de alertas. Com a crescente quantidade de componentes conectados na rede, os IDS tradicionais não estão sendo suficientes para a efetiva detecção de ataques maliciosos, tanto pelo volume de dados como pela crescente complexidade de novos ataques. Nesse sentido, a construção de uma arquitetura Internet Early Warning Systems (IEWS) possibilita detectar precocemente as ameaças, antes de causar algum perigo para os recursos da rede. O IEWS funciona como um coletor de diferentes geradores de alertas, possivelmente IDS, centralizando e correlacionado informações afim de gerar uma visão holística da rede. Sendo assim, o trabalho tem como objetivo descrever uma arquitetura IEWS para a correlação de alertas gerados por IDS dispersos geograficamente utilizando a técnica Case-Based Reasoning (CBR) em conjunto com Georreferenciamento de endereços IP. Os resultados obtidos nos experimentos, realizados sobre a implementação da técnica desenvolvida, mostraram a viabilidade da técnica na redução de alertas classificados como falsos-positivos. Isso demonstra a aplicabilidade da proposta como base para o desenvolvimento de técnicas mais apuradas de detecção dentro da arquitetura de IEWS estendida.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / INTRODUCTION. Adverse events have been the focus of studies in various countries, because they occur frequently and increase the morbidity and mortality of patients, constituting a new public health problem, directly impacting on patient safety. High alert medication (HAM) are more predictable to cause significant harm to the patient, even when used as intended. Other authors claim that the damage related to the HAM's lead not only suffering to the patient, but also raise the additional costs associated with care. OBJECTIVE. Evaluate the safety profile of HAM. METHODOLOGY. It was conducted active search for information through COCHRANE databases, LILACS, SciELO, SCOPUS, PubMed / MEDLINE and WEB OF SCIENCE. Then two reviewers independently conducted a preliminary evaluation of relevant titles, abstracts and then finally full text. RESULTS The initial search in the databases identified 1717 articles, which after exclusion of repeated work identified 427 titles, of which 53 are considered potentially relevant abstracts were selected. After evaluation of these, 25 were selected to read the full text. The systematic review evaluated seven articles, which showed that only 11 MAV's were identified in the literature could have serious events. The most frequently cited were warfarin (22.2%), cyclophosphamide (22.2%) and cyclosporine (22.2%).CONCLUSIONS. There were no reported any kind of error in the use of AVM's and particularly serious risks are unlikely because of the risks reported in patients related to drug classes involved. According to the criteria of quality, this revision came just seven articles that address the universe of MAV's, achieving high level of evidence only for six of these, which involved only some drugs: morphine, M6G-glicurônio, haloperidol, promethazine, ivabradine , digoxin, warfarin, ximelagatran, cyclophosphamide, cyclosporin and ATG. The review showed that one of the strategies used to improve security in hospitalized patients and use of AVM's is the creation of these medications protocol, and it is important to evaluate among the classes the drug that causes less damage. / INTRODUÇÃO. Os eventos adversos a medicamentos (EAM) têm sido foco de estudos em vários países, pois ocorrem com frequência e aumentam a morbimortalidade dos pacientes, constituindo-se em novo problema de saúde pública, impactando diretamente na segurança dos pacientes. Medicamentos de Alta Vigilância (MAV) são mais comuns de provocar danos significativos ao paciente, mesmo quando utilizados da forma prevista. Outros autores afirmam, que os danos relacionados aos MAV’s levam não só o sofrimento ao paciente, mas também elevam os custos adicionais associados com o cuidado. OBJETIVO. Avaliar o perfil de segurança dos MAV’s. METODOLOGIA. Foi realizado busca ativa de informações por meio das bases de dados COCHRANE, LILACS, SCIELO, SCOPUS, PUBMED/MEDLINE e WEB OF SCIENCE. Em seguida dois revisores de forma independente conduziram a avaliação inicial de títulos relevantes, posteriormente resumos e por fim texto completo. RESULTADOS. A busca inicial nas bases de dados identificou 1717 artigos, que após exclusão dos trabalhos repetidos identificou 427 títulos, dos quais foram selecionados 53 resumos considerados potencialmente relevantes. Após a avaliação destes, 25 foram selecionados para leitura do texto completo. A revisão sistemática avaliou sete artigos, os quais mostraram que apenas 11 MAV’s foram identificados na literatura com potenciais eventos graves. Os mais citados foram a varfarina (22,2%), ciclofosfamida (22,2%) e ciclosporina (22,2%). CONCLUSÕES. Não foram relatados nenhum tipo de erro no uso de MAV’s e notadamente riscos graves são pouco relatados provavelmente por conta da gravidade dos riscos em pacientes que utilizaram as classes medicamentosas envolvidas. De acordo com os critérios de qualidade, a referida revisão chegou a apenas sete artigos que abordam o universo dos MAV’s, conseguindo evidência de grau elevado apenas para seis destes, que envolveram somente alguns medicamentos: morfina, M6G-glicurônio, haloperidol, prometazina, ivabradina, digoxina, varfarina, ximelagatran, ciclofosfamida, ATG e ciclosporina. A revisão mostrou que uma das estratégias utilizadas para melhorar segurança em pacientes internados e em uso de MAV’s é a criação de protocolo de uso desses medicamentos, bem como é importante avaliar dentre as classes o medicamento que cause menos dano.

Made available in DSpace on 2016-12-23T14:37:40Z (GMT). No. of bitstreams: 1 Willian Bucker Moraes.pdf: 2543567 bytes, checksum: 39cbc60151124897818f6ebbce433b24 (MD5) Previous issue date: 2009-07-30 / The Rust, caused by the fungus Puccinia psidii Winter, is one of the most important diseases to the eucalypt host. The pathogen causes the disease in clonal minigarden and in young plants in the field, mostly in leaves and in young shoots. The favorable climatic conditions to infection of these pathogen in eucalypt include temperature between 18 to 25 ºC, with periods of at least 6 hours of leaf wetness for 5 to 7 days. Considering the interaction between environment and pathogen, the present study has as objective : (a) to map risk areas for the establishment of eucalyptus rust by exploiting the principle of space-time escape and (b) evaluate the potential impacts of global climate change on the spatial distribution of areas of risk for the occurrence of eucalyptus rust. (a) For the mapping of risk areas, it was calculated the rate of infection from the average daily maximum temperature and leaf wetness, and from this generated the risk index. The data used were obtained from the meteorological database, of the Regional Aracruz and São Mateus (Espírito Santo) and Regional Teixeira de Freitas (Bahia) for the years 2001 to 2006, a total of 23 weather stations. Based on the value of the index of occurrence risk of the eucalyptus rust, were prepared maps of spatial and temporal distribution of the disease, for which it was observed that the risk index varied according to the studied area and months of the year. The months from May to November were more favorable for the occurrence of eucalyptus rust. Thus, the results obtained in this study allowed the working principle of space-time escape, making it possible to schedule the harvest season, driving regrowth and planting of clones according to the level of resistance to disease. (b) To study the potential impacts of global climate change on the distribution of eucalypt rust, were prepared monthly maps of risk areas for the occurrence of disease, considering the current climatic conditions, based on a historical series from 1961 to 1990 and future scenarios A2 and B2, for the decades 2020, 2050 and 2080 provided by the Intergovermental Panel on Climate Change (IPCC). The weather conditions were classified into three categories, according to the potential risk of disease occurrence, considering the temperature (T) and relative humidity (RH): ): i) high risk (18 &#8804; T &#8804; 25 ºC and UR &#8805; 90%); ii) medium risk (T &#61500; 18 ou T > 25 ºC e UR 90%; T < 18 ou T &#61502; 25 ºC e UR &#8805; 90%); and iii) low risk (T 18 ou T > 25 ºC e UR 90%). Data about the future climate scenarios were provided by the GCM Change Fields. It was used in this work the simulation model Hadley Center for Climate Prediction and Research (HadCM3) using the Idrisi 32. Based on the results obtained observed that there will be reduction in the area favorable for the occurrence of eucalyptus rust, and this reduction will be gradual for the decades of 2020, 2050 and 2080 being more pronounced in scenario A2 than in B2. However, it is important to note that large areas still remain favorable for disease development, especially in the colder months of the year, in other words, from May to July. Thus, the knowledge generated in this work, coupled with the development of predictive models of the disease, can be important tools in the integrated management of eucalyptus rust / A ferrugem, causada pelo fungo Puccinia psidii Winter, é um das doenças mais importantes para a eucaliptocultura. O patógeno causa doença em minijardim clonal e em plantas novas no campo, principalmente em folhas e em brotações jovens. As condições climáticas favoráveis para infecção deste patógeno em eucalipto incluem temperatura entre 18 a 25 °C, com períodos de pelo menos 6 horas de molhamento foliar, por 5 a 7 dias consecutivos. Considerando a interação entre ambiente e patógeno, o presente trabalho teve como objetivos: (a) mapear áreas de risco ao estabelecimento da ferrugem do eucalipto explorando o princípio de escape espaço-temporal; e (b) avaliar os potenciais impactos das mudanças climáticas globais sobre a distribuição espacial das áreas de risco para ocorrência da ferrugem do eucalipto. (a) Para o mapeamento das áreas de risco, foi calculado o índice de infecção, a partir da média diária da temperatura máxima e do período de molhamento foliar, sendo a partir deste gerado o índice de risco. Os dados utilizados foram obtidos do banco de dados meteorológicos, das Regionais Aracruz e São Mateus (Espírito Santo) e da Regional Teixeira de Freitas (Bahia), referentes aos anos de 2001 a 2006, totalizando 23 estações meteorológicas. Com base no valor do índice de risco de ocorrência da ferrugem do eucalipto, elaboraram-se mapas de distribuição espaço-temporal da doença, pelos quais foi possível observar que o índice de risco variou em função da área estudada e dos meses do ano. Os meses de maio a novembro apresentaram maior favorabilidade para ocorrência da ferrugem do eucalipto. Sendo assim, os resultados obtidos nesse estudo permitiram trabalhar o princípio de escape espaço-temporal, tornando possível a programação da época de colheita, condução de rebrota e plantio de clones de acordo com o nível de resistência à doença. (b) Para o estudo dos impactos potenciais das mudanças climáticas globais sobre a distribuição da ferrugem do eucalipto, elaboraram-se mapas mensais das áreas de risco para ocorrência da doença, considerando as condições climáticas atuais, com base em uma série histórica de 1961 a 1990 e os cenários futuros A2 e B2, para as décadas de 2020, 2050 e 2080 disponibilizados pelo Intergovermental Panel on Climate Change (IPCC). As condições climáticas foram classificadas em três categorias, de acordo com o risco potencial de ocorrência da doença, considerando a temperatura (T) e umidade relativa do ar (UR): i) alto risco (18 &#8804; T &#8804; 25 ºC e UR &#8805; 90%); ii) médio risco (T &#61500; 18 ou T > 25 ºC e UR 90%; T < 18 ou T &#61502; 25 ºC e UR &#8805; 90%); e iii) baixo risco (T 18 ou T > 25 ºC e UR 90%). Os dados sobre os cenários climáticos futuros foram fornecidos pelo GCM Change Fields. Empregou-se neste trabalho o modelo de simulação Hadley Centre for Climate Prediction and Research (HadCm3), utilizando o software Idrisi 32. Com base nos resultados obtidos observou-se que haverá redução da área favorável para ocorrência da ferrugem do eucalipto, sendo que esta redução será gradativa para as décadas de 2020, 2050 e 2080, sendo mais acentuada no cenário A2 que no B2. Entretanto, é importante ressaltar que extensas áreas ainda continuarão favoráveis ao desenvolvimento da doença, principalmente nos meses mais frios do ano, ou seja, maio a julho. Desta forma, os conhecimentos gerados neste trabalho, aliados com o desenvolvimento de modelos de previsão da doença, podem constituir ferramentas importantes no manejo integrado da ferrugem do eucalipto

Orientadores: Luiz Henrique Antunes Rodrigues, Eduardo Delgado Assad / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Agricola / Made available in DSpace on 2018-08-12T21:35:45Z (GMT). No. of bitstreams: 1 Bucene_LucianaCorpas_D.pdf: 2595416 bytes, checksum: 86c930f5cf0a3ca7ba03de8acb811ea8 (MD5) Previous issue date: 2008 / Resumo: As perdas que ocorrem na agricultura são grandes, devido, principalmente, à ocorrência de sinistros climáticos que ocorrem nas plantações. Muitas vezes, os impactos social e econômico causados pelos danos são significativos, uma vez que envolvem fatores como a produção e o preço de alimentos. Como exemplos, têm-se a produção de café e a de cana-de-açúcar no Estado de São Paulo, que sofrem alternâncias motivadas por eventos climáticos adversos e, em especial, as geadas e as secas, que reduzem drasticamente as produções. Neste sentido, este estudo propõe identificar relações entre parâmetros climáticos, como temperatura máxima, temperatura mínima, precipitação, entre outros atributos, visando descobrir eventuais novos conhecimentos, a partir do comportamento conhecido dos atributos climáticos já ocorridos no passado, com o propósito de desenvolver a previsão local de geada e a previsão de deficiência hídrica. Para isso, foram aplicadas técnicas de descoberta de conhecimento em grandes bancos de dados climáticos. Utilizaram-se as ferramentas WEKA e o DISCOVER, que foram consideradas satisfatórias, uma vez que os objetivos propostos foram atingidos. As bases de dados disponíveis atenderam a necessidade para a realização do projeto, apresentando um volume de dados e atributos suficientes para que pudesse gerar resultados para a previsão local de geada e de deficiência hídrica. Referente aos resultados, com até 1 dia de antecedência à geada, o modelo gerado foi considerado confiável. A partir de 2 dias de antecedência à geada, os resultados encontrados apresentam uma diminuição no grau de acerto quanto mais distante estiver de acontecer o evento geada. Para o caso deficiência hídrica, os resultados encontrados foram diferenciados conforme a classe. Para a classe não, com 1dia até 15 dias de antecedência ao evento, o grau de acerto foi alto e aceitável. A classe forte, em seguida à classe não, é a que apresenta melhores resultados de acerto, decaindo para as outras classes. Até 3 dias de antecedência ao evento deficiência hídrica e, dependendo do mês, o grau de acerto é aceitável. De 4 dias em diante, os resultados mostram que o modelo gerado não é aceitável / Abstract: The losses that occur in agriculture are high, mainly due to the occurrence of crop damages due to climatic events. Many times, the social and economic impacts caused by the damages are significant, since they involve factors such as the production and the price of foods. For example, coffee and sugarcane production in São Paulo State suffer alternations motivated by adverse climatic events and, in special, frost and drought, that greatly reduce the production. The purpose of this study is to identify relationships between climatic parameters, such as maximum temperature, minimum temperature, precipitation, etc., in order to discover eventual new knowledge, from known behavior of the climatic attributes already occurred in the past, with the objective of developing local frost and deficit water forecast models. To achieve this, data mining techniques were applied to climatic data bases. WEKA and the DISCOVER tools had been used and considered satisfactory, since they reached the objectives. The available databases were suitable for the accomplishment of the project, presenting enough volume of data and attributes so that it could generate results for the frost and water deficit forecast. Concerning to the results, with up to 1 day of antecedence to the frost, the generated model was considered trustworthy. From 2 days of antecedence to the frost the results present a reduction in the accuracy. For water deficit, results were differentiated, depending on the class. For the not class, from 1 to 15 days of antecedence to the event, the accuracy was high and acceptable. The strong class, following the not class, is the one that presents better results, falling down for the other classes. Up to 3 days of antecedence to the event water deficit and, depending on the month, the accuracy is acceptable. For 4 days or more in advance, the results showed that the generated model is not acceptable / Doutorado / Doutor em Engenharia Agrícola