論文摘要 所得稅為政府的一項重要經常性收入，也是眾多公共支出的來源，行政部門每年編製預算書時，均須作審慎合理的估測，以符精確表達之要求。準此，本文研究目的有二：第一，預估所得稅稅收之合理收入數。第二個目的，如何分配予各區國稅局，以落實預算之執行及績效之考核。 在運用單一ARIMA，以稅收實徵數之月資料來預測綜合所得稅及營利事業所得稅之稅收，其有效模式分別為(1,0,1)及(1,1,1)，值得參考運用。在迴歸模式上面，綜合所得稅稅收預測與國民所得之解釋變數有正向關係，惟經檢定殘差有自我相關之現象，因此將之修正以AR(1) 誤差來作預測迴歸式，相較所編列預算數，修正後迴歸式較精確;營利事業所得稅之稅收預測，經逐步迴歸結果，與國內生產毛額之解釋變數有正向關係，而以變數取log之迴歸式有較佳之效果。 就管理層面而言，稅收預算之執行須配合事後績效評估與考核，惟先決條件須稅收預算數合理分配予五區國稅局。經實證分析結果，以國民所得與國內生產毛額之解釋變數來估計其綜合所得稅及營利事業所得稅之預算數，不論在全國所得稅稅收或分五區估計上，其差異性不大，且相當合理可信，值得各稽徵單位分配稅收預算數之參考。

Time series forecasting and modeling are challenging problems during the past decades, because of its plenty of properties and underlying correlated relationships. As a result, researchers proposed a lot of models to deal with the time series. However, the proposed models such as Autoregressive integrated moving average (ARIMA) and artificial neural networks (ANNs) only describe part of the properties of time series. In this thesis, we introduce a new hybrid model integrated filter structure to improve the prediction accuracy. Case studies with real data from University of Kentucky HealthCare are carried out to examine the superiority of our model. Also, we applied our model to operating room (OR) to reduce the inefficiency cost. The experiment results indicate that our model always outperforms compared with other models in different conditions.

This master thesis is focused on analysis and forecasting of hourly and daily electricity price on the deregulated Czech daily electricity market. The methods used for estimating and forecasting hourly and daily prices are picked from the ARIMA-GARCH family of models and Neural Networks. For daily price data, the Redundant Haar Wavelet Transform decomposition of the time series is used in combination with ARIMA and Neural Networks models for forecasting. For hourly data, ARIMA and Neural Network models are considered. The forecasting results of daily data indicate that simpler models such as seasonal ARIMA outperform all other methods. Also the wavelet decomposi- tion of the daily series didn't prove useful in enhancing the forecast precision. For hourly data, the Multilayer Perceptron architecture of the neural network outperformed the ARIMA forecast. JEL Classification C20, C22, C45, C53, C65 Keywords Forecasting, Time Series, ARIMA, GARCH, Neural Net- works, Wavelet Transform Author's e-mail krystof.cerny@gmail.com Supervisor's e-mail lebovicm@gmail.com 1

n the electrical power grid, the power load is not constant but continuouslychanging. This depends on many different factors, among which the habits of theconsumers, the yearly seasons and the hour of the day. The continuous change inenergy consumption requires the power grid to be flexible. If the energy provided bygenerators is lower than the demand, this is usually compensated by using renewablepower sources or stored energy until the power generators have adapted to the newdemand. However, if buffers are depleted the output may not meet the demandedpower and could cause power outages. The currently adopted practice in the indus-try is based on configuring the grid depending on some expected power draw. Thisanalysis is usually performed at a high level and provide only some basic load aggre-gate as an output. In this thesis, we aim at investigating techniques that are able topredict the behaviour of loads with fine-grained precision. These techniques couldbe used as predictors to dynamically adapt the grid at run time. We have investigatedthe field of time series forecasting and evaluated and compared different techniquesusing a real data set of the load of the Swedish power grid recorded hourly throughyears. In particular, we have compared the traditional ARIMA models to a neuralnetwork and a long short-term memory (LSTM) model to see which of these tech-niques had the lowest forecasting error in our scenario. Our results show that theLSTM model outperformed the other tested models with an average error of 6,1%.

Antibiotic resistant bacteria, a growing health crisis, arise due to antibiotic overuse and misuse. Resistant infections endanger the lives of patients and are financially burdensome. Aggregate antimicrobial susceptibility reports, called antibiograms, are critical for tracking antibiotic susceptibility and evaluating the likelihood of the effectiveness of different antibiotics to treat an infection prior to the availability of patient specific susceptibility data. This research leverages the Massachusetts Statewide Antibiogram database, a rich dataset composed of antibiograms for $754$ antibiotic-bacteria pairs collected by the Massachusetts Department of Public Health from $2002$ to $2016$. However, these antibiograms are at least a year old, meaning antibiotics are prescribed based on outdated data which unnecessarily furthers resistance. Our objective is to employ data science techniques on these antibiograms to assist in developing more responsible antibiotic prescription practices. First, we use model selectors with regression-based techniques to forecast the current antimicrobial resistance. Next, we develop an assistant to immediately identify clinically and statistically significant changes in antimicrobial resistance between years once the most recent year of antibiograms are collected. Lastly, we use k-means clustering on resistance trends to detect antibiotic-bacteria pairs with resistance trends for which forecasting will not be effective. These three strategies can be implemented to guide more responsible antibiotic prescription practices and thus reduce unnecessary increases in antibiotic resistance.

Antibiotic resistant bacteria, a growing health crisis, arise due to antibiotic overuse and misuse. Resistant infections endanger the lives of patients and are financially burdensome. Aggregate antimicrobial susceptibility reports, called antibiograms, are critical for tracking antibiotic susceptibility and evaluating the likelihood of the effectiveness of different antibiotics to treat an infection prior to the availability of patient specific susceptibility data. This research leverages the Massachusetts Statewide Antibiogram database, a rich dataset composed of antibiograms for $754$ antibiotic-bacteria pairs collected by the Massachusetts Department of Public Health from $2002$ to $2016$. However, these antibiograms are at least a year old, meaning antibiotics are prescribed based on outdated data which unnecessarily furthers resistance. Our objective is to employ data science techniques on these antibiograms to assist in developing more responsible antibiotic prescription practices. First, we use model selectors with regression-based techniques to forecast the current antimicrobial resistance. Next, we develop an assistant to immediately identify clinically and statistically significant changes in antimicrobial resistance between years once the most recent year of antibiograms are collected. Lastly, we use k-means clustering on resistance trends to detect antibiotic-bacteria pairs with resistance trends for which forecasting will not be effective. These three strategies can be implemented to guide more responsible antibiotic prescription practices and thus reduce unnecessary increases in antibiotic resistance.

The aim of this thesis is to apply and evaluate potential forecasting models for solar power production, based on data from a photovoltaic facility in Sala, Sweden. The thesis evaluates single step forecasting models as well as multiple step forecasting models, where the three compared models for single step forecasting are persistence, autoregressive integrated moving average (ARIMA) and ARIMAX. ARIMAX is an ARIMA model that also takes exogenous predictors in consideration. In this thesis the evaluated exogenous predictor is wind speed. The two compared multiple step models are multiple step persistence and the Gaussian process (GP). Root mean squared error (RMSE) is used as the measurement of evaluation and thus determining the accuracy of the models. Results show that the ARIMAX models performed most accurate in every simulation of the single step models implementation, which implies that adding the exogenous predictor wind speed increases the accuracy. However, the accuracy only increased by 0.04% at most, which is determined as a minimal amount. Moreover, the results show that the GP model was 3% more accurate than the multiple step persistence; however, the GP model could be further developed by adding more training data or exogenous variables to the model.

Os processos espaço-temporais vêm ganhando destaque nos últimos anos, em razão do aumento de estudos compreendendo variáveis que apresentam interação entre as dimensões espacial e temporal. Com o objetivo de modelar esses processos, Pfeifer e Deutsch (1980a) propuseram uma extensão da classe de modelos univariados de Box-Jenkins, denominada por modelo espaço-temporal autoregressivo de média móvel (STARMA). Essa classe de modelos é utilizada para descrever dados de séries temporais espacialmente localizadas. Os processos passíveis de modelagem via classe de modelos STARMA são caracterizados por observações de variáveis aleatórias, em que os locais a serem incorporados no modelo são fixos no espaço. A dependência entre as n séries temporais é modelada por meio da matriz de ponderação, de modo que os modelos da classe STARMA expressem cada observação no tempo t e na localização i como uma média ponderada de combinações lineares das observações anteriores e a inovação defasada no espaço e no tempo conjuntamente. Dada a nova classe de modelos, os objetivos deste estudo foram apresentar a classe de modelos STARMA, implentar computacionalmente, no software R, rotinas que permitam a análise de dados espaço-temporais, com as rotinas implementadas estabelecer e testar modelos de séries temporais aos dados de temperaturas mínimas médias mensais de 8 estações meteorológicas situadas no Paraná e comparar a classe de modelos STARMA com a classe de modelos univariados proposta por Box e Jenkins (1970). Com este estudo verificou-se que na apresentação da classe de modelos STARMA há complexidade no conceito de ordens de vizinhança e na identificação dos modelos espaço-temporais. Em relação a criação de rotinas responsáveis pelas análises de dados espaço-temporais observou-se dificuldades em sua implementação, principalmente no momento de estimação dos parâmetros. Na comparação da classe de modelos STARMA, multivariada, com a classe de modelos SARIMA, univariada, constatou-se que ambos os modelos se ajustaram de maneira satisfatória aos dados, produzindo previsões acuradas. / Spatio-temporal processes have been highlighted lately, due to the increase of studies approaching variables that present interactions between the spatial and temporal dimensions. In order to model these processes, Pfeifer e Deutsch (1980a) have suggested an extension of the Box-Jenkins univariate model class, named spatio-temporal autoregressive moving-average model (STARMA). This model class is used to describe spatially located time series data. The processes prone to be modeled via the STARMA model class are characterized by observations of random variables, in which the locations to be incorporated in the model are spatially fixed. The dependence between the n time series is modeled through the weighing matrix. So STARMA models express each observation at time t and location i as a weighed mean of linear combinations of the previous observations and the jointly lagged innovation in space and time. Given the new class models, the objectives of this study were to present a class of models STARMA, implentar computationally, in textit R software, routines that allow the analysis of spatio-temporal data with the routines implemented to establish and test models time series data of monthly average minimum temperatures of 8 meteorological stations located in Paraná and compare the class of models STARMA with the class of univariate models proposed by Box e Jenkins (1970). With this study it was found that the presentation of the class of models STARMA no complexity in the concept of ordered neighborhood and identification of spatio-temporal models. Regarding the creation of routines responsible for the analysis of spatio-temporal observed difficulties in its implementation, especially at the time of estimation of parameters. In comparison class STARMA models, multivariate, with the class of SARIMA models, univariate, it was found that both models were adjusted satisfactorily to the data, producing accurate forecasts.

In this thesis, player’s performance on ice hockey is modelled to create newmetricsby match and season for players. AD-trees have been used to summarize ice hockey matches using state variables, which combine context and action variables to estimate the impact of each action under that specific state using Markov Decision Processes. With that, an impact measure has been described and four player metrics have been derived by match for regular seasons 2007-2008 and 2008-2009. General analysis has been performed for these metrics and ARIMA models have been used to analyze and predict players performance. The best prediction achieved in the modelling is the mean of the previous matches. The combination of several metrics including the ones created in this thesis could be combined to evaluate player’s performance using salary ranges to indicate whether a player is worth hiring/maintaining/firing

The purpose of this report is to predict annual insurance data with quarterly data as predictors and to evaluate its accuracy against other naive prediction models. A relationship is discerned between the two data categories and the interest goes beyond publication frequency as there is a fundamental difference between quarterly and annual data. The insurance industry organization Insurance Sweden publishes quarterly data that contain all insurance events reported while the annual data only contain insurance events which led to disbursement from the insurance companies. This discrepancy shows to be problematic when predicting annual outcomes. Forecasts are estimated by ARIMA models on short time series and compared with classic linear regression models. The implied results from all insurance subcategories in traffic, motor vehicles and household- and corporate insurance are that, in some cases, prediction using linear regression on quarterly data is more precise than the constructed naive prediction models on annual data. However, the results vary between subcategories and the regression models using quarterly data need further improvement before it is the obvious choice when forecasting annual number of events that led to disbursements from the insurance companies.