Orientador: Percival Bueno de Araujo / Banca: Laurence Duarte Colvara / Banca: Mara Lúcia Martins Lopes / Banca: Luis Filomeno de Jesus Fernandes / Banca: Igor Kopcak / Resumo: Neste trabalho é proposta a utilização da rede neural artificial (RNA) ARTMAP Nebulosa (fuzzy) no ajuste de parâmetros de controladores suplementares para o amortecimento de oscilações eletromecânicas de sistemas elétricos de potência, visando tornar este ajuste mais eficiente. Análises comparativas da atuação das redes neurais artificiais ARTMAP Nebulosa e Perceptron Multicamadas (PM) são realizadas para dois sistemas multimáquinas considerando o ajuste individual e coordenado dos controladores. Tais redes são utilizadas para o projeto dos controladores ESP (Estabilizadores de Sistemas de Potência) e POD (Power Oscillation Damping) acoplado ao dispositivo FACTS (Flexible Alternating Current Transmission Systems) UPFC (Unified Power Flow Controller). Será evidenciado que a RNA ARTMAP Nebulosa pode ser utilizada na melhora da estabilidade dinâmica, fornecendo resultados muito semelhantes aos da RNA Perceptron Multicamadas. Entretanto, é importante enfatizar que a vantagem da utilização da RNA ARTMAP Nebulosa está no fato da garantia da estabilidade e plasticidade associadas a um rápido treinamento, o que não ocorre com a RNA Perceptron Multicamadas / Abstract: This work proposes the use of artificial neural network (ANN) Fuzzy ARTMAP to adjust the parameters of additional controllers to damp electromechanical oscillations in electric power systems in order to make this adjustment more efficient due to variations in load. Comparative analysis of the performance of artificial neural networks Fuzzy ARTMAP and Multilayer Perceptron are performed for two multimachine systems, considering individual and coordinated controller adjustment. Those networks are used for the design of Power System Stabilizers (PSS) and Power Oscillation Damping (POD) that is coupled to the FACTS (Flexible Alternating Current Transmission Systems) UPFC (Unified Power Flow Controller). It will be shown that the ANN Fuzzy ARTMAP can be used in the improvement of dynamic stability, providing very similar results to the ANN Multilayer Perceptron. However, it is important to emphasize that the advantage of using ANN Fuzzy ARTMAP is the guarantee of stability and plasticity associated with a fast training process which does not occur for the ANN Multilayer Perceptron / Doutor

Garantir a qualidade da carne bovina produzida no Brasil tem sido uma preocupação dos produtores, pois contribui para aumentar a exportação e o consumo interno do produto. Por isso, tem-se pesquisado novos métodos que analisam e garantam a qualidade da carne, de forma rápida, eficiente e não destrutiva. A ressonância magnética nuclear (RMN) tem se destacado como uma das técnicas de controle de qualidade de carne. Neste trabalho as redes neurais artificiais estão sendo utilizadas para o reconhecimento de padrões dos dados de ressonância magnética nuclear oriundos de carne bovina. Mais especificamente, os respectivos dados têm sido utilizados por uma rede perceptron multicamadas para a extração de características da carne bovina, possibilitando a classificação do grupo genético e do sexo dos animais a partir de uma amostra da referida carne. Os resultados dos experimentos são também apresentados para ilustrar o desempenho da abordagem proposta. / Guaranteeing the quality of the bovine meat produced in Brazil has been a concern of the producers because it contributes to increase the export and the domestic consumption of the product. Therefore, new methods have been researched that analyze and guarantee the quality of the meat in a fast, efficient and non destructive way. Nuclear magnetic resonance (NMR) has been highlighted as one of the techniques of meat quality control. In this work study artificial neural networks are being used for pattern recognition from data obtained by the resonance equipment, originating from bovine meat. More specifically, the respective data have been used by a multilayer perceptron network for extraction of bovine meat characteristics, making possible the classification of both genetic group and animal sex starting from a single meat sample. Several results of experimental tests are also presented to illustrate the performance of the proposed approach.

A suinocultura é uma atividade de grande importância em termos mundiais e de Brasil. Entretanto, por serem animais homeotérmicos, algumas alterações no ambiente térmico de alojamento podem alterar suas respostas fisiológicas e comportamentais para manutenção da temperatura interna. Portanto, o objetivo dessa pesquisa foi avaliar o comportamento alimentar de suínos, mediante a influência do ambiente térmico, nas fases de crescimento e terminação para diferentes linhagens comerciais e sexo. Além disso, buscou-se o desenvolvimento de modelos matemáticos e sistemas inteligentes para predição do tempo em alimentação (TM, min dia-1) dos suínos. Os dados foram coletados em uma granja experimental de suínos, localizada na cidade de Clay Center, Nebraska, Estados Unidos. O período experimental contemplou duas estações durante o ano 2015/2016 (verão e inverno), totalizando 63 dias (9 semanas) de informações coletadas para cada estação. Os animais alojados foram de três linhagens comerciais distintas: Landrace, Duroc e Yorkshire. Cada baia apresentava composição mista, sendo alojados 40 animais de diferentes linhagens comerciais e sexo. No total, foram confinados 240 animais, sendo 80 animais para cada linhagem comercial entre machos castrados e fêmeas. Foram registrados dados de temperatura do ar (Tar, °C), temperatura do ponto de orvalho (Tpo, °C) e umidade relativa do ar (UR, %) a cada 5 minutos no interior da instalação. Para TM, os dados foram coletados e registrados a cada 20 segundos por meio de um sistema de coleta de dados por rádio frequência. O conforto térmico foi analisado a partir do Índice de Temperatura e Umidade (ITU) e a Entalpia Específica (H, kJ kg-1 de ar seco). Para avaliar a relação entre o ambiente térmico e TM, foi utilizada estatística multivariada por meio de análise de componentes principais (ACP) e agrupamento para obtenção de padrões e seleção de variáveis para entrada nos modelos. O modelo fuzzy e as redes neurais artificias foram desenvolvidos em ambiente MATLAB® R2015a por meio dos toolboxes Fuzzy e Neural Network, com o objetivo de predizer TM, tendo como variáveis de entrada: linhagem comercial, sexo, idade e ITU. De uma maneira geral, as médias de Tar estiveram dentro da zona de termoneutralidade (ZCT) em todo período experimental, sendo que apenas a UR apresentou valores abaixo da UR crítica inferior. Para o ITU, apenas no verão foram encontrados valores acima da ZCT, entretanto, esses valores estiveram abaixo do ITU crítico superior. Diante da análise dos resultados, pôde-se observar em relação ao comportamento alimentar, que a fêmea Landrace apresentou o menor tempo em alimentação com médias de 42,19 min dia-1 e 43,73 min dia-1 para o inverno e verão, respectivamente, seguido do macho castrado de mesma linhagem. Enquanto as demais linhagens apresentaram valores acima de 60 min dia-1. Não foi observado correlação linear significativa entre o ambiente térmico e TM uma vez que os animais estiveram dentro de sua ZCT ao longo de todo período experimental, indicando que o comportamento alimentar foi influenciado principalmente pelos fatores homeostáticos e cognitivos-hedônicos. A estatística multivariada dividiu os animais em 8 grupos. Foi observado que animais de linhagens e sexos distintos se comportaram da mesma maneira, dificultando a modelagem matemática. Entretanto, alguns grupos apresentaram maior quantidade de animais de determinada linhagem e sexo, sendo estes utilizados como \"grupos padrão\" para o desenvolvimento do modelo fuzzy e a rede neural artificial. O modelo fuzzy apresentou R2 de 0,858 quando utilizado os dados do grupo padrão, entretanto, para todos os valores o R2 foi de 0,549. Já a rede neural apresentou um R2 de 0,611 para os dados completos e R2 de 0,914 para o \"grupo padrão\". Portanto, a rede neural artificial mostrou-se como uma ferramenta de maior precisão e acurácia na predição do comportamento alimentar de suínos nas fases de crescimento e terminação. / The swine production in an activity of great importance to Brazil and to the world. However, because they maintain a constant body temperature and, alterations in the thermic accommodation environment can directly affect their physiological and behavioral responses for maintaining the internal temperature. Thus, the objective of this study was to access the feeding behavior of growing-finishing pigs of different sirelines and gender and its relationship with climate variables (thermic environment). Furthermore, mathematical models based on classic logic was developed as well as an intelligent system for predicting the total time spent eating (TM, min day -1). The data was collected in an experimental farm located in Clay Center, Nebraska, United States. The experimental period contemplated two seasons (summer and winter), totalizing 63 days (9 weeks) of information collected for each season. The housed animals were from three different commercial sirelines: Landrace, Duroc and Yorkshire. Each pen presented a mix composition, being housed 40 animals of different sirelines and gender. In total, there were 240 housed animals, being 80 animals for each sireline among barrows and gilts. The data registered were air temperature (Tar, °C), dew point temperature (Tpo, °C) and relative humidity of the air (UR, %) every 5 minutes inside the facility. For TM, the data were collected and registered every 20 seconds by a radio frequency data collection system. The thermal comfort was analyzed from the Temperature and Humidity Index (THI) and Specific Enthalpy (H, kJ kg-1 of dry air). In order to evaluate the relationship between the thermic environment and TM, the multivariate statistics through principal component analysis (PCA) and grouping was utilized for obtaining the selection standards of variables to enter in the models. The fuzzy model and the artificial neural networks were developed in a MATLAB® R2015a environment through the Fuzzy and the Neural Network toolboxes with the objective to predict TM, having as entry variables: sireline, gender, age and THI. On the whole, the Tar averages were inside the thermoneutral zone (ZCT), however, these values were below the superior critic THI. In the face of the results analysis, it could be observed in ration to the feeding behavior that the Landrace gilt presented the shortest time eating with averages of 42.19 min day-1 and 43.73 min day-1 for winter and summer respectively followed by the barrow from the same sireline, while the other sirelines presented values above 60 min day-1. It was not observed a significative linear correlation between the thermic environment and TM once the animals were inside their ZCT throughout all the experimentation period, indicating that the feeding behavior was influenced mainly by the homeostatic and cognitivehedonic factors. The multivariate statistics divided the animals in 8 groups, being observed that animals of different sirelines and gender behave the same way throughout the experimentation period, making the mathematical modeling difficult. However, some groups presented a bigger amount of animals of determined sireline and gender, being utilized as \"standard groups\" for the development of the fuzzy model and the artificial neural network. The fuzzy model presented an R2 of 0,858 when utilizing the \"standard group\" data, however, for all the values the R2 was 0.549. In the other hand the neural network presented an R2 of 0.611 for the complete data and an R2 of 0.914 for the \"standard group\". Thus, the artificial neural network appeared to be a tool of a better precision and accuracy when predicting the feeding behavior of pigs on growing-finishing phases.

Este trabalho apresenta um sistema de trading que toma decisões de compra e de venda do índice Standard & Poors 500, na modalidade seguidor de tendência, mediante o emprego de redes neurais artificiais multicamadas com propagação para frente, no período de 5 anos, encerrado na última semana do primeiro semestre de 2012. Geralmente o critério usual de escolha de redes neurais nas estimativas de preços de ativos financeiros é o do menor erro quadrático médio entre as estimativas e os valores observados. Na seleção das redes neurais foi empregado o critério do menor erro quadrático médio na amostra de teste, entre as redes neurais que apresentaram taxas de acertos nas previsões das oscilações semanais do índice Standard & Poors 500 acima de 60% nessas amostras de teste. Esse critério possibilitou ao sistema de trading superar a taxa anual de retorno das redes neurais selecionadas pelo critério usual e, por larga margem, a estratégia de compre e segure no período. A escolha das variáveis de entrada das redes neurais recaiu entre as que capturaram o efeito da anomalia do momento dos preços do mercado de ações no curto prazo, fenômeno amplamente reconhecido na literatura financeira. / This work presents a trend follower system that makes decisions to buy and sell short the Standard & Poors 500 Index, by using multilayer feedforward neural networks. It was considered a period of 5 years, ending in the last week of the first half of 2012. Usually a neural networks choice criterion to forecast financial asset prices is based on the least mean square error between the estimated and observed prices in the test samples. In this work we also adopted another criterion based on the least mean square error for those neural networks that had a hit rate above 60% of the Standard & Poors 500 Index weekly change in the test sample. This criterion was shown to be the most appropriate one. The neural networks input variables were chosen among those technical indicators that better captured the anomaly of the short term momentum of prices. The annual rate of return of the trading system based on those criteria surpassed those selected by the usual criteria, and by a wide margin the buy-and-hold strategy. The neural networks inputs were chosen to capture the momentum anomaly of the prices on the short term that is fully recognized in the financial literature.

Three dimensional (3D) ground reaction force (GRF) are an essential component for gait analysis. Current methods for measuring 3D GRF involve using a stationary force plate embedded in the ground, which captures the forces as subjects walk across the platform. This approach has several limitations, a few being: it can only capture a few steps at a time, it is expensive to purchase and maintain, it can't reflect forces caused by natural uneven surfaces, etc. Previous research has attempted to develop wearable force sensors to overcome these problems; however, these endeavors have resulted in devices that are expensive, bulky, and fail to accurately measure forces when compared to static force plates. This thesis presents the implementation and validation of novel nanocomposite piezoresponsive foam (NCPF) sensors for measuring 3D GRF. Four NCPF sensors were embedded in a shoe sole at four locations: heel, arch, ball, and toe. The signals from each sensor were used in a functional data analysis (FDA) to develop a statistical model for estimating 3D GRF. The process of calibrating the sensors to model GRF was validated through a study where 9 subjects (4 females, 5 males) walked on a force-sensing treadmill for two minutes. Two approaches were used to model the GRF response. The first approach was based on functional decomposition of the data. Using a tenfold cross validation process a statistical model was developed for each subject with the ability to predict walking 3D GRF with less than 7% error. The second approach used machine learning to model 3D GRF. Using the same walking data for the statistical models, an artificial neural network (ANN) was used to create subject-specific models that could predict walking 3D GRF with less than 11% error. The predictive capabilities of ANN were tested using a pilot study where a single subject performed a calibration procedure by running at seven different speeds for thirty seconds each on the force-sensing treadmill. This calibration data was used to train a model, which was then used to estimate vertical GRF (VGRF) for three additional running trials at randomly selected speeds from within the calibration range. The ANN model was able to predict VGRF for three running speeds after calibration with less than 4% error. The use of NCPF sensors to estimate 3D GRF was shown to be a viable alternative to static force plates. It is recommended, in future work, that 3D GRF and subsequent sensor data be collected from a large sample of subjects to create a baseline of 3D GRF characteristics for a population that will enable a robust cross-subject model capable of performing real-time ground reaction force analysis across the general population, which will greatly benefit our understanding of human gait.

Extreme Learning Machine (ELM) is a recently discovered way of training Single Layer Feed-forward Neural Networks with an explicitly given solution, which exists because the input weights and biases are generated randomly and never change. The method in general achieves performance comparable to Error Back-Propagation, but the training time is up to 5 orders of magnitude smaller. Despite a random initialization, the regularization procedures explained in the thesis ensure consistently good results. While the general methodology of ELMs is well developed, the sheer speed of the method enables its un-typical usage for state-of-the-art techniques based on repetitive model re-training and re-evaluation. Three of such techniques are explained in the third chapter: a way of visualizing high-dimensional data onto a provided fixed set of visualization points, an approach for detecting samples in a dataset with incorrect labels (mistakenly assigned, mistyped or a low confidence), and a way of computing confidence intervals for ELM predictions. All three methods prove useful, and allow even more applications in the future. ELM method is a promising basis for dealing with Big Data, because it naturally deals with the problem of large data size. An adaptation of ELM to Big Data problems, and a corresponding toolbox (published and freely available) are described in chapter 4. An adaptation includes an iterative solution of ELM which satisfies a limited computer memory constraints and allows for a convenient parallelization. Other tools are GPU-accelerated computations and support for a convenient huge data storage format. The chapter also provides two real-world examples of dealing with Big Data using ELMs, which present other problems of Big Data such as veracity and velocity, and solutions to them in the particular problem context.

Orientador: Marcelo Libânio / Resumo: O objetivo do presente trabalho é avaliar modelos de previsão de tarifa de água, aplicados a autarquias municipais e empresas privadas, nas regiões Sul e Sudeste do Brasil. Utilizando a metodologia de cálculo e posterior comparação dos erros obtidos para as previsões, verificando também a aplicabilidade das tarifas previstas para cada sistema de abastecimento. Utilizou-se dois modelos de previsão, o primeiro, fundamentado em técnicas de regressão linear múltipla e o segundo, baseado na aplicação de redes neurais artificiais. Avaliando, dessa forma, a capacidade de os dois modelos em questão preverem os valores tarifários a serem cobrados pelos prestadores de serviços de abastecimento de água e coleta de esgoto, a partir da análise das tarifas anteriormente praticadas. Os dados subsidiários para elaboração dos modelos foram obtidos por meio do sistema nacional de informações sobre saneamento (SNIS). Confirmada a consistência do banco de dados primário, procedeu-se com processamento destes dados, e definição das variáveis mais intervenientes para a definição da tarifa por meio da técnica de análise de correlação. Propôs-se a classificação dos sistemas de acordo com a classe jurídica do prestador de serviço, os cenários financeiros (superávit ou déficit) destes prestadores e o porte populacional dos municípios atendidos. Os resultados obtidos indicaram que os processos de previsão, em ambos os modelos utilizados, foram capazes de prever com elevada acurácia as tarifas, e garanti... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The objective of the present work was evaluating forecasting models for water tariff applied to municipal and private companies in the South and Southeast regions of Brazil. Using the calculation methodology and subsequent comparison of the errors obtained for the forecasts, also verifying the applicability of the forecast tariffs for each supply system. Two prediction models are used, the first based on multiple linear regression techniques and the second based on the application of artificial neural networks. Evaluating, in this way, the ability of the two models in question to predict the tariff values to be charged by the water supply and wastewater collection service providers, based on the analysis of the tariffs previously practiced. The subsidiary data for the elaboration of the models were obtained through the national sanitation information system (SNIS). Confirming the consistency of the primary database, we proceeded with processing of these data and definition of the most intervening variables for the definition of the tariff through the correlation analysis technique. The classification of the systems according to the legal class of the service provider, the financial scenarios (surplus or deficit) of these providers and the population size of the municipalities served were proposed. The obtained results indicated that the forecasting processes, in both models used, were able to predict with high accuracy the tariffs, and guaranteed the maintenance of the surplu... (Complete abstract click electronic access below) / Mestre

A cana-de-açúcar é uma importante cultura semi-perene em regiões tropicais do mundo como a principal fonte de açúcar e bioenergia e o Brasil é seu maior produtor. Como qualquer outra cultura, demanda um aperfeiçoamento prática constante, buscando uma cultura sustentável e com maiores rendimentos e menores custos. Uma das alternativas é a utilização de práticas de agricultura de precisão para explorar a variabilidade espacial dos rendimentos potenciais e para tanto, os mapas de produtividade são essenciais. Para obter os dados necessários para gerar um mapa confiável, é necessário um sistema com capacidade de ler e georreferenciar os dados do sensor e compará-los a uma calibração. No entanto, os resultados das pesquisas mais recentes associadas aos monitores de rendimento comercial, que utilizam apenas um tipo de sensor para determinar os mapas de produtividade, não retratam a exatidão exigida para a cana-de-açúcar. Este estudo teve como objetivo explorar o potencial do uso de dados provenientes de sensores instalados em diferentes partes da colhedora de cana-de-açúcar para determinação e aplicação em monitores de produtividade e determinação de falhas na lavoura. Para fins de comparação foi utilizado um transbordo instrumentado com células de carga para aferição da massa colhida. Foram utilizadas abordagens estatísticas convencionais e inteligência artificial para fusão dos dados e predição da produtividade da cana-de-açúcar, os métodos convencionais foram regressão linear simples e múltipla, e comparado com o método de redes neurais. Além da produtividade foi possível constatar que é possível identificar as falhas na lavoura através dos dados coletados e das falhas produzidas manualmente, todos os sensores medidos identificaram as falhas georeferenciadas. Com relação aos modelos implementados e utilizados, os baseados em regressão linear múltipla não apresentaram potencial na integração e predição da produtividade com os valores de erros definidos nas premissas do trabalho que é de menor que 2%. Além disso os mapas gerados com esses modelos tiveram algumas discrepâncias quanto ao aumento da produtividade em algumas áreas e extração das falhas existentes. Já o modelo de fusão utilizando redes neurais artificiais apresentou uma excelente alternativa para predição da produtividade. Uma vez que a rede é treinada, a mesma apresentou erros inferiores a 2% em todos os mapas gerados. De maneira geral todos os sensores quando avaliados individualmente apresentaram vantagens e desvantagens na determinação da produtividade. Porém, quando fundido os dados dos diversos sensores, as respostas encontradas apresentaram coeficiente de determinação R2 superiores a 95%, RMSE menor que 1kg e RE menor que 2%. / Sugarcane is an important semi-perennial crop in tropical regions of the world as the main source of sugar and bioenergy, and Brazil is its largest producer. Like any other culture, it demands constant improvement in practice, seeking a sustainable culture with higher yields and lower costs. One of the alternatives is the use of precision farming practices to explore the spatial variability of potential yields and for that, productivity maps are essential. To obtain the data needed to generate a reliable map, a system is required that is capable of reading and georeferencing the sensor data and comparing them to a calibration. However, the results of the most recent surveys associated with commercial yield monitors, which use only one type of sensor to determine productivity maps, do not depict the exactitude required for sugarcane. This study aimed to explore the potential of using data from sensors installed in different parts of the sugarcane harvester for determination and application in productivity monitors and determination of crop failure, for comparison purposes a transhipment was used instrumented with load cells to measure the mass harvested. We used conventional statistical approaches and artificial intelligence for data fusion and prediction of sugarcane productivity, conventional methods were simple and multiple linear regression, and compared with the neural network method. In addition to productivity, it was possible to verify that it is possible to identify crop failures through the data collected and the failures produced manually, all the measured sensors identified georeferenced faults. Regarding the implemented and used models, those based on multiple linear regression did not present potential in the integration and prediction of productivity with the values of errors defined in the assumptions of the work that is less than 2%. In addition, the maps generated with these models had some discrepancies regarding productivity increase in some areas and extraction of existing flaws. On the other hand, the model of fusion using artificial neural networks presented an excellent alternative for prediction of productivity; since the network is trained the same one presented in all the generated maps errors inferior to 2%. In a general way all the sensors when evaluated individually presented advantages and disadvantages in determining the productivity, but when fused the data of the various sensors the answers found of coefficient of determination R2 higher than 95%, RMSE less than 1kg and RE less than 2%.

[Truncated abstract] The identification of individual animals based on acoustic parameters is a non-invasive method of identifying individuals with considerable advantages over physical marking procedures. One requirement for an effective and practical method of acoustic individual identification is that it is call-independent, i.e. determining identity does not require a comparison of the same call or song type. This means that an individuals identity over time can be determined regardless of any changes to its vocal repertoire, and different individuals can be compared regardless of whether they share calls. Although several methods of acoustic identification currently exist, for example discriminant function analysis or spectrographic cross-correlation, none are call-independent. Call-independent identification has been developed for human speaker recognition, and this thesis aimed to: 1) determine if call-independent identification was possible in birds, using similar methods to those used for human speaker recognition, 2) examine the impact of noise in a recording on the identification accuracy and determine methods of removing the noise and increasing accuracy, 3) provide a comparison of features and classifiers to determine the best method of call-independent identification in birds, and 4) determine the practical limitations of call-independent identification in birds, with respect to increasing population size, changing vocal characteristics over time, using different call categories, and using the method in an open population. ... For classification, Gaussian mixture models and probabilistic neural networks resulted in higher accuracy, and were simpler to use, than multilayer perceptrons. Using the best methods of feature extraction and classification resulted in 86-95.5% identification accuracy for two passerine species, with all individuals correctly identified. A study of the limitations of the technique, in terms of population size, the category of call used, accuracy over time, and the effects of having an open population, found that acoustic identification using perceptual linear prediction and probabilistic neural networks can be used to successfully identify individuals in a population of at least 40 individuals, can be used successfully on call categories other than song, and can be used in open populations in which a new recording may belong to a previously unknown individual. However, identity was only able to be determined with accuracy for less than three months, limiting the current technique to short-term field studies. This thesis demonstrates the application of speaker recognition technology to enable call-independent identification in birds. Call-independence is a pre-requisite for the successful application of acoustic individual identification in many species, especially passerines, but has so far received little attention in the scientific literature. This thesis demonstrates that call-independent identification is possible in birds, as well as testing and finding methods to overcome the practical limitations of the methods, enabling their future use in biological studies, particularly for the conservation of threatened species.

本研究首次提出以未來臨界報酬率為輸出變數，利用兩種類神經網路(Artificial Neural Network)估算國內電子股代表樣本報酬率的風險值(Value at Risk , VaR)。在研究設計上考慮到使用不同期長來計算自變項所帶來的影響而產生兩種預測方法。本研究並以回顧檢定(Backtesting )檢討藉由臨界值報酬率作為類神經估計法與一般以變異數/共變數法或蒙地卡羅模擬法所估算出VaR的差異。 綜合本研究，在學術及實務上的貢獻有下列四點： 1. 設計臨界報酬率作為估算VaR的方式，可以避免以往計算VaR時，報酬率分配主觀給定的問題。 2. 相關研究過去並未同時涉及類神經網路與VaR，而本研究首次應用類神經網路估算VaR。 3. 本文亦提出以多種不同的基本變數衡量期長來估算VaR，或可幫助界定差異的研究設計。 4. 本研究使用類神經網路可能的一項限制是報酬率臨界值 的設計方式；而類神經網路可能勝出其它預測工具的理由可能是 (1)學習到隱性因子的特性 (2)預測方式為非線性 (3)毋須依賴常態或特定分配之假設。以往類神經網路研究在賽馬決定各工具優劣時，較少探究類神經勝出或落敗的理由，而這卻是本研究設計的焦點。