nÃo hà / O presente trabalho tem como objetivo principal realizar um estudo sobre o spread bancÃrio no Brasil durante o perÃodo de 2011 a 2014, analisando os seus principais determinantes e construindo um modelo de previsÃo. Foi utilizado o modelo economÃtrico dos MÃnimos Quadrados OrdinÃrios (MQO) e, adicionalmente, foi feito um modelo de previsÃo do spread. A base de dados utilizada foi extraÃda do Banco Central do Brasil (BCB): spread mÃdio das operaÃÃes de crÃdito, InadimplÃncia da carteira de crÃdito (pessoa fÃsica e jurÃdica), recolhimentos obrigatÃrios de instituiÃÃes financeiras e o endividamento das famÃlias com o Sistema Financeiro Nacional. Dentre outros resultados obtidos, verificou-se que caso ocorra um aumento de 1% da inadimplÃncia do crÃdito de pessoas fÃsicas haverà um aumento de 0,38% do Spread mÃdio no Brasil. Por outro lado, uma elevaÃÃo de 1% da inadimplÃncia de crÃdito de pessoa jurÃdica elevarà o spread mÃdio em 0,63%. Sem dÃvida, isso mostra a importÃncia dessas duas variÃveis na determinaÃÃo do spread mÃdio pelos bancos brasileiros. Por outro lado, o modelo de previsÃo utilizado permitiu concluir que do perÃodo de dezembro de 2014 a junho de 2015 o Spread mÃdio serà de aproximadamente 13% em Janeiro; 12,66% em Fevereiro; 12,63% em marÃo; 12,74% em Abril e 12,87% em Maio e 13,14% em junho de 2015. / This paper aims to conduct a study on the banking spread in Brazil during the period from 2011 to 2014, analyzing its main determinants and building a predictive model. We used the econometric model of Ordinary Least Squares (OLS) and additionally was done a spread prediction model. The database used was taken from the Central Bank of Brazil (BCB): average spread of loans, Bad debt loan portfolio (individual and corporate), reserve requirements of financial institutions and the household debt to the National Financial System. Among other results, it was found that if an increase of 1% of physical persons of credit default there will be an increase of 0.38% of the average spread in Brazil. On the other hand, an increase of 1% of corporate credit default raise the average spread 0.63%. No doubt, this shows the importance of these two variables in determining the average spread for Brazilian banks. On the other hand the forecasting model concluded that the period December 2014 to June 2015, it was found that during this period the average Spread will be approximately 13% in January; 12.66% in February; 12.63% in March; 12.74% in April and 12.87% in May to 13.14% in June 2015.

The development of the Severe Weather Impact Forecasting System (SWIFS) for flash flood hazards in South Africa is described in this thesis. Impact forecasting addresses the need to move from forecasting weather conditions to forecasting the consequential impact of these conditions on people and their livelihoods. SWIFS aims to guide disaster managers to take early action to minimise the adverse effects of flash floods focussing on hotspots where the largest impact is expected. The first component of SWIFS produced an 18-hour probabilistic outlook of potential occurrence of flash floods. This required the development of an ensemble forecast system of rainfall for small river basins (the forecasting model component), based on the rainfall forecast of a deterministic numerical weather prediction model, to provide an 18-hour lead-time, taking into account forecast uncertainty. The second component of SWIFS covered the event specific societal and structural impacts of these potential flash floods, based on the interaction of the potential occurrence of flash floods with the generalised vulnerability to flash floods of the affected region (the impact model component). The impact model required an investigation into the concepts of regional vulnerability to flash floods, and the development of relevant descriptive and mathematical definitions in the context of impact forecasting. The definition developed in the study links impact forecasting to the likelihood and magnitude of adverse impacts to communities under threat, based on their vulnerability and due to an imminent severe weather hazard. Case studies provided evidence that the concept of SWIFS can produce useful information to disaster managers to identify areas most likely to be adversely affected in advance of a hazardous event and to decide on appropriate distribution of their resources between the various hotspots where the largest impacts would be. SWIFS contributes to the current international research on short-term impact forecasting by focussing on forecasting the impacts of flash floods in a developing country with its limited spatial vulnerability information. It provides user-oriented information in support of disaster manager decision-making through additional lead-time of the potential of flash floods, and the likely impact of the flooding. The study provides a firm basis for future enhancement of SWIFS to other severe weather hazards in South Africa. / Thesis (PhD)--University of Pretoria, 2015. / gm2015 / Geography, Geoinformatics and Meteorology / PhD / Unrestricted

Le travail que nous présentons dans cette thèse porte sur le problème d'affectation dans une place de marché de prospects Internet. Plus précisément, ce travail a pour ambition de répondre à la problématique de l'adéquation de l'offre et de la demande, dans un contexte caractérisé par des flux continus faisant évoluer en temps réel l'ensemble des offres disponibles et les demandes à satisfaire. Pour ce faire, nous proposons dans un premier temps un modèle mono-période qui optimise le problème d'affectation à un instant donné et en considérant une seule période de temps, tout en permettant la prise en compte instantanée des nouvelles offres et demandes et leur adéquation en temps réel. Ce modèle permet d'optimiser deux objectifs à savoir: la maximisation du chiffre d'affaires et la satisfaction des clients.Par la suite nous proposons d'étendre ce modèle sur plusieurs périodes de temps futures afin de prendre en compte l'aspect temps réel de l'activité de la place de marché et donc le fait que des flux continus font évoluer en temps réel l'ensemble des offres et des demandes. L'objectif étant de tirer profit de la connaissance concernant cette évolution, par le biais de l'intégration d'un modèle de prévision dans un modèle d'optimisation multi-période.Ainsi, nous proposons un modèle d'optimisation multi-période permettant d'envisager à un instant donné des affectations sur plusieurs périodes de temps futures afin de réaliser les meilleures affectations possibles. Aussi, nous proposons un modèle de prévision des nouveaux flux tout en considérant les caractéristiques du modèle d'optimisation multi-période.Construire un modèle de prévision nécessite de définir les données à prévoir avant d'envisager toute méthode de prévision. En d'autres termes, nous devons choisir les paramètres du modèle de prévision, à savoir: les données historiques appropriées, le pas de temps de la prévision ainsi que l'horizon de la prévision. Le défi consiste donc à définir les paramètres du modèle de prévision qui conviendront au fonctionnement du modèle de l'optimisation multi-période.Par ailleurs, une des caractéristiques de la place de marché est la temporalité de son système. Ainsi, nous proposons un algorithme assurant l'aspect temps réel et donc le fait que les affectations s'effectuent toutes les minutes. L'algorithme que nous proposons fonctionne de manière continue à longueur de journée en optimisant à chaque instant l'adéquation offre/demande de prospects Internet tout en considérant instantanément les flux continus de prospects Internet ainsi que la mise à jour régulière de la demande Enfin, pour mettre en évidence l'efficacité et les bénéfices que la place de marché peut en tirer par l'utilisation des modèles et de l'algorithme proposés, nous avons mené des tests et différentes expérimentations sur des données réelles. Ces tests nous ont permis de valider nos travaux et d'évaluer la qualité des résultats obtenus.L'objectif de ce travail est double, d'une part, donner un cadre solide et formel pour répondre à la problématique de la place de marché de prospects Internet. D'autre part, le cadre proposé devrait être aussi générique que possible afin de résoudre tout autre problème analogue à celui de la place de marché de prospects Internet. / The work that we present in this thesis focuses on the assignment problem in a marketplace of Internet prospects. More precisely, this work aims to address the problem of matching offers and demands in a context characterized by a continuous flows. These latter evolve inreal time the set of available offers and demands to satisfy. To do this, we propose initially a mono-period model which optimizes the assignment problem at a given instant and taking into account asingle period of time while allowing the instantaneous consideration of new offers and demands and their adequacy in real time. This model considers two objectives to optimize, namely: maximization of turnover as well as clients satisfaction.Thereafter, we propose to extend this model over several future time periods in order to take into account the real time aspect of the marketplace activity and so the fact that a continuous flows evolve in real time the set of offers en demands. The objective is to take advantage of knowledge about this evolution, through the integration of a forecasting model in a multi-period optimization model. Thus,we propose a multi-period optimization model for considering at agiven instant assignments over several future time periods. Also, we propose a forecasting model for new flows while considering the characteristics of the multi-period optimization model.Building a forecasting model requires defining the data before considering any forecasting method. In other words, we have to choose the parameters of the forecasting model, namely the appropriate historical data, the forecasting time step and the forecasting horizon. The challenge is to define the parameters of the forecasting model which agree with the functioning the multi-period optimization model.Furthermore, a feature of the marketplace is the temporality of its system. Thus, we propose an algorithm ensuring real-time aspect and so the fact that assignments are made every minute. The proposed algorithm works continuously all day long while optimizing every instant the offer/demand adequacy of Internet prospects and instantly considering the continuous flux of Internet prospects as well as the regular updating demand. Finally, in order to show the efficiency and the benefits that the marketplace can reap by the use of the proposed models, we conducted tests and various experiments on real data. These tests have allowed us to validate the proposed models and evaluate the quality of the results.The aim is twofold, giving a strong and formal framework to address the issue of the marketplace of Internet prospects but also proposing a generic framework to solve any problem similar to that of the marketplace of Internet prospects.

In the current study, we have presented a systematic analysis of the diurnal cycle of rainfall over the Indian region using satellite observations, and evaluated the ability of the Weather Research and Forecasting Model (WRF) to simulate some of the salient features of the observed diurnal characteristics of rainfall. Using high resolution simulations, we also investigate the underlying mechanisms of some of the observed diurnal signatures of rainfall. Using the Tropical Rain-fall Measuring Mission (TRMM) 3-hourly, 0.25 ×0.25 degree 3B42 rainfall product for nine years (1999-2007), we extract the finer spatial structure of the diurnal scale signature of Indian summer monsoon rainfall. Using harmonic analysis, we construct a signal corresponding to diurnal and sub-diurnal variability. Subsequently, the 3-hourly time-period or the octet of rain-fall peak for this filtered signal, referred to as the “peak octet,” is estimated with care taken to eliminate spurious peaks arising out of Gibbs oscillations. Our analysis suggests that over the Bay of Bengal, there are three distinct modes of the peak octet of diurnal rainfall corresponding to 1130, 1430 and 1730 IST, from north central to south Bay. This finding could be seen to be consistent with southward propagation of the diurnal rainfall pattern reported by earlier studies. Over the Arabian sea, there is a spatially coherent pattern in the mode of the peak octet (1430 IST), in a region where it rains for more than 30% of the time. In the equatorial Indian Ocean, while most of the western part shows a late night/early morning peak, the eastern part does not show a spatially coherent pattern in the mode of the peak octet, owing to the occurrence of a dual maxima (early morning and early/late afternoon). The Himalayan foothills were found to have a mode of peak octet corresponding to 0230 IST, whereas over the Burmese mountains and the Western Ghats (west coast of India) the rainfall peaks during late afternoon/early evening (1430-1730 IST). This implies that the phase of the diurnal cycle over inland orography (e.g., Himalayas) is significantly different from coastal orography (e.g., Western Ghats). We also find that over the Gangetic plains, the peak octet is around 1430 IST, a few hours earlier compared to the typical early evening maxima over land. The second part of our study involves evaluating the ability of the Weather Research and Fore-casting Model (WRF) to simulate the observed diurnal rainfall characteristics. It also includes conducting high resolution simulations to explore the underlying physical mechanisms of the observed diurnal signatures of rainfall. The model (at 54km resolution) is integrated for the month of July 2006 since this period was particularly favourable for the study of diurnal cycle. We first evaluate the sensitivity of the model to the prescribed sea surface temperature (SST) by using two different SST datasets, namely Final Analyses (FNL) and Real-time Global (RTG). The overall performance of RTG SST was found to be better than FNL, and hence it was used for further model simulations. Next, we investigated the impact of different parameterisations (convective, microphysical, boundary layer, radiation and land surface) on the simulation of diurnal cycle of rainfall. Following this sensitivity study, we identified the suite of physical parameterisations in the model that “best” reproduces the observed diurnal characteristics of Indian monsoon rainfall. The “best” model configuration was used to conduct two nested simulations with one-way, three-level nesting (54-18-6km) over central India and Bay of Bengal. While the 54km and 18km simulations were conducted for July 2006, the 6km simulation was carried out for the period 18-24 July 2006. This period was chosen for our study since it is composed of an active period (19-21 July 2006), followed by a break period (22-24 July 2006). At 6km grid-spacing the model is able to realistically simulate the active and break phases in rainfall. During the chosen active phase, we find that the observed rainfall over central India tends to reach a maximum in the late night/early morning hours. This is in contrast to the observed climatological diurnal maxima of late evening hours. Interestingly, the 6km simulation for the active phase is able to reproduce this late night/early morning maxima. Upon further analysis, we find that this is because of the strong moisture convergence at the mid-troposphere during 2030-2330 IST, leading to the rainfall peak seen during 2330-0230 IST. Based on our analysis, we conclude that during both active and break phases of summer monsoon, mid-level moisture convergence seems to be one of the primary factors governing the phase of the diurnal cycle of rainfall. Over the Bay of Bengal, the 6km model simulation is in very good agreement with observations, particularly during the active phase. The southward propagation observed during 19-20 July 2006, which was not captured by the coarse resolution simulation (54km), is exceedingly well captured by the 6km simulation. The positive anomalies in specific humidity attain a maxima during 2030-0230 IST in the north and during 0830-1430 IST in the south. This confirms the role of moisture convergence in the southward propagation of rainfall. Equally importantly we find that while low level moisture convergence is dominant in the north Bay, it is the mid-level moisture convergence that is predominant in the south Bay.

Made available in DSpace on 2014-08-20T14:25:48Z (GMT). No. of bitstreams: 1 dissertacao_dionis_blank.pdf: 3213477 bytes, checksum: bb5a4cb38573c6480453829c0fd6cd2b (MD5) Previous issue date: 2009-03-06 / The Antarctic Sea Ice Concentration (ASIC) is thought to be an important element in the analysis of the world climate. However, few studies have investigated its relation to other climatic elements. Thus, the aim of this study was to verify the relation between the ASIC and low temperatures in South America through two approaches. The first, regional, investigated the occurrence of a connection between the ASIC and the cold and hot quantiles of the daily lowest temperature as observed in some weather stations in Rio Grande do Sul in the 1982 2005 period. For such, low temperature values were transformed into cold and hot quantiles through the quantile technique, and correlated to ASIC sectors. The correlation coefficient showed a connection between the elements, with emphasis on the influence of Weddell, Ross Sea sectors and Bellingshausen and Amundsen Sea sector, especially because the Indian Ocean and the Western Pacific Ocean are farther away. The second approach, continental, analyzed the ASIC variability and its connection with low temperatures observed in South America by means of NCEP-NCAR reanalysis in the 1982 2007 period. For such, the sectors of larger ASIC variability were identified through the principal component analysis technique, enabling the adjustment of the ASIC-based low temperature forecasting model to South America to the data set by the model and the observed data in the reanalysis through the multiple lineal regression analysis technique. The prevailing areas for the explanation of ASIC variability were found to be in the sectors above mentioned. The worst (best) adjustment of the model occurred in the cold (hot) period, when there is a greater (smaller) variability of low temperatures and smaller (greater) ASIC variability. / A Concentração de Gelo Marinho Antártico (CGMA) é considerada um elemento importante na análise do clima mundial. Contudo, poucos estudos têm investigado sua relação com outros elementos climáticos. Desse modo, o objetivo deste trabalho consistiu em verificar a relação entre a CGMA e a temperatura mínima na América do Sul utilizando duas abordagens. Na primeira, regional, examinou-se a existência de conexão entre a CGMA e as classes fria e quente da temperatura mínima diária, observada em algumas estações meteorológicas do Rio Grande do Sul, no período de 1982 a 2005. Para isso, os dados de temperatura mínima foram transformados em classes fria e quente, por meio da técnica dos quantis, e correlacionados com os setores da CGMA. O coeficiente de correlação mostrou a existência de conexão entre os elementos, com destaque para a influência dos setores dos Mares de Weddell, de Ross e de Bellingshausen e Amundsen, até porque os setores do Oceano Índico e do Oceano Pacífico Oeste apresentam maior distância. Na segunda abordagem, continental, analisou-se a variabilidade da CGMA e sua ligação com a temperatura mínima na América do Sul, observada pela reanálise do NCEP-NCAR, no período de 1982 a 2007. Para isso, os setores de maior variabilidade da CGMA foram identificados por intermédio da técnica de análise de componentes principais, possibilitando o ajuste de um modelo de previsão de temperatura mínima para a América do Sul, baseado na CGMA, com dados previstos pelo modelo e dados observados pela reanálise, mediante o uso da técnica de análise de regressão linear múltipla. As áreas mais predominantes na explicação da variabilidade da CGMA foram encontradas nos setores já citados. O pior (melhor) ajuste do modelo ocorreu no período frio (quente), onde existe maior (menor) variabilidade da temperatura mínima e menor (maior) variabilidade da CGMA.

Thesis is aimed on creation of prediction model for releasing medium-term water stream flow forecasts. Created model create forecasts based on principal of finding most similar historical case. Usefulness of forecasting model is demonstrated for operation of one isolated reservoir in gauge profile Oslavany on river Oslava.

No description available.

協同規劃、預測與補貨(Collaborative Planning, Forecasting and Replenishment,CPFR)，在歐美經過一些企業的採用後已經有顯著的成效，目前國內已經有一些企業相繼採用或即將採用CPFR，期望能因此降低供應鏈作業成本及提升供應鏈作業績效，以提升企業競爭力。在CPFR流程與供應鏈協同作業環境下，一個供需雙方協同，且績效良好的的銷售預測具有關鍵的重要性，是管理決策與協同合作時的的重要依據；但是多數的企業並沒有一個結構化、系統化的預測流程及方法，而是各部門透過簡單時間序列方法、天真預測法或人為經驗法則估算需求，進行多點且不同方法之預測，這樣的銷售預測較無穩定的品質，亦較難提供管理者合理的數據解釋。本研究結合時間序列、多元回歸模型與基因演算法發展出一個CPFR流程下之三階段混合預測方法，以買賣方直接之銷售資料、銷售計畫等資訊進行以「週」為單位之個別商品銷售預測。同時本研究中，亦以國內某製造業公司與其顧客（一國際大型零售連鎖店通路商）之產品銷售資料進行方法的驗證；實驗顯示，本研究所提出之預測方法之預測結果較Jeong等人(2002)所提結合多元回歸模型與基因演算法之二階段預測系統之預測結果佳；亦較傳統使用普通最小平方法求解之一般統計回歸方法預測結果佳。 / It has been verified in pilot projects by many European and American Corporations that Collaborative Planning, Forecasting and Replenishment (CPFR) can improve supply chain performance. Enterprises nowadays in Taiwan are implementing or going to implement CPFR, with hopes to reduce their supply chain operation cost, enhance logistic performance and increase their competition capability consequently. Under CPFR process and supply chain collaboration environment, a supply and demand both sides promised identical sales forecast with well forecasting performance for order decision making and cooperation is very important. Due to the dynamic complexities of both internal and external co-operate environment, many firms resort to qualitative, navie forecasting or other simple quantitative forecasting techniques and have many forecasts in their organization. However, these forecasting techniques lack the structure and extrapolation capability of quantitative forecasting models or without stable performance, while multi-forecasts providing different views of demand. Forecasting inaccuracies exist and typically lead to dramatic disturbances in sales order and production planning. This paper presents a hybrid forecasting model for sales forecasting requirements in CPFR. A three stage model is proposed that integrate the time series model, regression model and use genetic algorithm to determine its coefficients efficiently. Direct sales information and related planned events in both collaborated sides is used for individual product’s “week” sales forecasting. To verify this model, we experiment on two different products and produce forecasts with datum from one manufacturer in Taiwan and its international retailer. The results shows that the hybrid sales forecasting model has better forecasting performance than not only the causal-genetic forecasting model proposed by Jeong et al. (2002), but also ordinary regression model with no genetic training process.

對投資者而言，投資股市的目的就是賺錢，但影響股價因素眾多，我們要如何判斷明天是漲是跌？因此如何建立一個準確的預測模型，一直是財務市場研究的課題之一，然而財務市場一直被認為是一個複雜.充滿不確定性及非線性的動態系統，這也是在建構模型上一個很大的阻礙，本篇研究中使用的EEMD方法則適合解決如金融市場或氣候等此類的非線性問題及有趨勢性的資料上。 在本研究中，我們將EEMD結合ANN建構出兩種不同形式的模型去進行台股個股的預測，也試圖改善ARMA模型使其預測效果較好；此外為了能夠達到分散風險的效果，採用了投資組合的方式，在權重的決定上，我們結合動態與靜態的方式來計算權重；至於在交易策略上，本研究也加入了移動平均線，希望能找到最適合的預測模型，本研究所使用的標的物為曾在該期間被列為注意股票的10檔股票。 另外，我們也分析了影響台股個股價格波動的因素，透過EEMD拆解，我們能夠從中得到具有不同意義的本徵模態函數(IMF)，藉由統計值分析重要的IMF其所代表的意義。例如：影響高頻波動的重要因素為新聞媒體或突發事件，影響中頻的重要因素為法人買賣及季報，而影響低頻的重要因素則為季節循環。 結果顯示，EEMD-ANN Model 1是一個穩健的模型，能夠創造出將近20%的年報酬率，其次為EEMD-ANN Model 2，在搭配移動平均線的策略後，表現與Model 1差不多，但在沒有配合移動平均線策略時，雖報酬率仍為正，但較不穩定，因此從研究結果也可以看到，EEMD-ANN的模型皆表現比ARMA的預測模型好。 / The main purpose of investing is to earn profits for an investor, but there are many factors that can influence stock price. Investments want to know the price will rise or fall tomorrow. Therefore, how to establish an accurate forecasting model is one of the important issue that researched by researchers of financial market. However, the financial market is considered of a complex, uncertainty, and non-linear dynamic systems. These characteristics are obstacles on constructing model. The measure, EEMD, used in this study is suitable to solve questions that are non-linear but have trends such as financial market, climate and so on. In this thesis, we used three models including ARMA model and two types of EEMD-ANN composite models to forecast the stock price. In addition, we tried to improve ARMA model, so a new model was proposed. Through EEMD, the fluctuation of stock price can be decomposed into several IMFs with different economical meanings. Moreover, we adopted portfolio approach to spread risks. We integrate the static weight and the dynamic weight to decide the optimal weights. Also, we added the moving average indicator to our trading strategy. The subject matters in this study are 10 attention stocks. Our results showed that EEMD-ANN Model 1 is a robust model. It is not only the best model but also can produce near 20% of 1-year return ratio. We also find that our EEMD-ANN model have better outcome than those of the traditional ARMA model. Owing to that, the increases of trading performance would be expected via the selected EEMD-ANN model.

觀光事業素有無煙囪工業之稱，自政府於民國四十八年全力推動發展以來 ，來華觀光旅客人數即不斷地成長，此對促進國民外交與增加政府的外匯 收入上有莫大的幫助。觀光旅客人數的多寡，直接影響本地觀光業者與政 府相關單位對觀光業軟硬體設施的投資，像是觀光旅館的興建、導遊人員 的培訓以及整體策略的規劃。不當的評估或不正確的需求預測，都將導致 大量觀光資源的閒置或浪費。本研究計劃主要應用簡算法、時間趨勢模式 、時間序列模式、計量經濟模式，尋找並建立來華觀光旅客長短期需求預 測模式，並針對總體或各主要市場的需求，利用各種模式評估準則提出一 最佳預測模式，以供政府相關單位與觀光業者作為往後政策釐定以及投資 計劃擬定時的參考。