The research concerning differences in duration of the Great Recession is limited and inconclusive. We define duration of crisis as the count of years lost due to the crisis, and estimate the determinants of crisis duration on the dataset of 54 developed and developing countries. This thesis contrasts with previous literature by employing Bayesian Model Averaging (BMA) to accommodate for the large amount of potential explanatory variables and to address model uncertainty. Moreover, an innovative measure of export competitiveness, which accounts for the changes in non-price factors such as quality, is used. The results bring suggestive evidence of positive impact of developed financial markets, high share of private consumption and improvements in export competitiveness. We also find positive effect of fiscal policy stimulus once it is controlled for the feedback loop of uncertainty which appears when heavily indebted countries finance fiscal stimulus through issuance of additional debt. Lastly, it needs to be concluded, that the results are not robust to all prior specifications. In particular, the more restrictive Beta binomial model prior shrinks the statistical significance of aforementioned results heavily. JEL Classification F12, F21, F23, H25, H71, H87 Keywords Great Recession, Crisis duration, Economic...

In this thesis different situations are considered in which the preliminary test estimator is applied and the performance of the preliminary test estimator under different proposed loss functions, namely the reflected normal , linear exponential (LINEX) and bounded LINEX (BLINEX) loss functions is evaluated. In order to motivate the use of the BLINEX loss function rather than the reflected normal loss or the LINEX loss function, the risk for the preliminary test estimator and its component estimators derived under BLINEX loss is compared to the risk of the preliminary test estimator and its components estimators derived under both reflected normal loss and LINEX loss analytically (in some sections) and computationally. It is shown that both the risk under reflected normal loss and the risk under LINEX loss is higher than the risk under BLINEX loss. The key focus point under consideration is the estimation of the regression coefficients of a multiple regression model under two conditions, namely the presence of multicollinearity and linear restrictions imposed on the regression coefficients. In order to address the multicollinearity problem, the regression coefficients were adjusted by making use of Hoerl and Kennard’s (1970) approach in ridge regression. Furthermore, in situations where under- or overestimation exist, symmetric loss functions will not give optimal results and it was necessary to consider asymmetric loss functions. In the economic application, it was shown that a loss function which is both asymmetric and bounded to ensure a maximum upper bound for the loss, is the most appropriate function to use. In order to evaluate the effect that different ridge parameters have on the estimation, the risk values were calculated for all three ridge regression estimators under different conditions, namely an increase in variance, an increase in the level of multicollinearity, an increase in the number of parameters to be estimated in the regression model and an increase in the sample size. These results were compared to each other and summarised for all the proposed estimators and proposed loss functions. The comparison of the three proposed ridge regression estimators under all the proposed loss functions was also summarised for an increase in the sample size and an increase in variance. / Thesis (PhD)--University of Pretoria, 2014. / lk2014 / Statistics / PhD / Unrestricted

Historians, mostly engaging with written evidence, have argued that the Christianisation of the Roman Empire resulted in changes in both attitudes and behaviour towards children, resulting in a decrease in their maltreatment by society. I begin with a working hypothesis that this attitude-change was real and resulted in a reduction in the maltreatment of children; and that this reduction in maltreatment is evident in the literature. The approach to investigating this hypothesis belongs to the emerging field of digital humanities: by using programming techniques developed in the field of sentiment analysis, I create two sentiment-analysis like tools, one a lexicon-based approach, the other an application of a naive bayes machine learning approach. The latter is favoured as more accurate. The tool is used to automatically tag sentences, extracted from a corpus of texts written between 100 B.C and 600 A.D, that mention children, as to whether the sentences feature the maltreatment of children or not. The results are then quantitively analysed with reference to the year in which the text was written, with no statistically significant result found. However, the high accuracy of the tool in tagging sentences, at above 88%, suggests that similar tools may be able to play an important role, alongside traditional research techniques, in historical and social-science research in the future.

In this paper, Natural Language Processing and classification algorithms were used to create a program that automatically can tag different errands that are connected to Fortnox (an IT company based in Växjö) support service. Controlled experiments were conducted to find the best classification algorithm together with different Bag-of-Word pre-processing algorithms to find what was best suited for this problem. All data were provided by Fortnox and were manually labeled with tags connected to it as training and test data. The result of the final algorithm was 69.15% correctly/accurately predicted errands using all original data. When looking at the data that were incorrectly predicted a pattern was noticed where many errands have identical text attached to them. By removing the majority of these errands, the result was increased to 94.08%.

Nowadays, the ideas of continuous integration and continuous delivery are under heavy usage in order to achieve rapid software development speed and quick product delivery to the customers with good quality. During the process ofmodern software development, the testing stage has always been with great significance so that the delivered software is meeting all the requirements and with high quality, maintainability, sustainability, scalability, etc. The key assignment of software testing is to find bugs from every test and solve them. The developers and test engineers at Ericsson, who are working on a large scale software architecture, are mainly relying on the logs generated during the testing, which contains important information regarding the system behavior and software status, to debug the software. However, the volume of the data is too big and the variety is too complex and unpredictable, therefore, it is very time consuming and with great efforts for them to manually locate and resolve the bugs from such vast amount of log data. The objective of this thesis project is to explore a way to conduct log analysis efficiently and effectively by applying relevant machine learning algorithms in order to help people quickly detect the test failure and its possible causalities. In this project, a method of preprocessing and clusering original logs is designed and implemented in order to obtain useful data which can be fed to machine learning algorithms. The comparable log analysis, based on two machine learning algorithms - Recurrent Neural Network and Naive Bayes, is conducted for detecting the place of system failures and anomalies. Finally, relevant experimental results are provided and analyzed.

Artificial intelligence (AI) agents will need to interact with both other AI agents and humans. One way to enable effective interaction is to create models of associates to help to predict the modeled agents' actions, plans, and intentions. If AI agents are able to predict what other agents in their environment will be doing in the future and can understand the intentions of these other agents, the AI agents can use these predictions in their planning, decision-making and assessing their own potential. Prior work [13, 14] introduced the S# algorithm, which is designed as a robust algorithm for many two-player repeated games (RGs) to enable cooperation among players. Because S# generates actions, has (internal) experts that seek to accomplish an internal intent, and associates plans with each expert, it is a useful algorithm for exploring intent, plan, and action in RGs. This thesis presents a graphical Bayesian model for predicting actions, plans, and intents of an S# agent. The same model is also used to predict human action. The actions, plans and intentions associated with each S# expert are (a) identified from the literature and (b) grouped by expert type. The Bayesian model then uses its transition probabilities to predict the action and expert type from observing human or S# play. Two techniques were explored for translating probability distributions into specific predictions: Maximum A Posteriori (MAP) and Aggregation approach. The Bayesian model was evaluated for three RGs (Prisoners Dilemma, Chicken and Alternator) as follows. Prediction accuracy of the model was compared to predictions from machine learning models (J48, Multi layer perceptron and Random Forest) as well as from the fixed strategies presented in [20]. Prediction accuracy was obtained by comparing the model's predictions against the actual player's actions. Accuracy for plan and intent prediction was measured by comparing predictions to the actual plans and intents followed by the S# agent. Since the plans and the intents of human players were not recorded in the dataset, this thesis does not measure the accuracy of the Bayesian model against actual human plans and intents. Results show that the Bayesian model effectively models the actions, plans, and intents of the S# algorithm across the various games. Additionally, the Bayesian model outperforms other methods for predicting human actions. When the games do not allow players to communicate using so-called cheaptalk, the MAP-based predictions are significantly better than Aggregation-based predictions. There is no significant difference in the performance of MAP-based and Aggregation-based predictions for modeling human behavior when cheaptalk is allowed, except in the game of Chicken.

Empirical Bayes methods have been around for a long time and have a wide range of applications. These methods provide a way in which historical data can be aggregated to provide estimates of the posterior mean. This thesis revisits some of the empirical Bayesian methods and develops new applications. We first look at a linear empirical Bayes estimator and apply it on ranking and symbolic data. Next, we consider Tweedie’s formula and show how it can be applied to analyze a microarray dataset. The application of the formula is simplified with the Pearson system of distributions. Saddlepoint approximations enable us to generalize several results in this direction. The results show that the proposed methods perform well in applications to real data sets.

Natural language processing has many important applications in today, such as translations, spam filters, and other useful products. To achieve these applications supervised and unsupervised machine learning models, have shown to be successful. The most important aspect of these models is what the model can achieve with different datasets. This article will examine how RNN models compare with Naive Bayes in text classification. The chosen RNN models are long short-term memory (LSTM) and gated recurrent unit (GRU). Both LSTM and GRU will be trained using the flair Framework. The models will be trained on three separate datasets with different compositions, where the trend within each model will be examined and compared with the other models. The result showed that Naive Bayes performed better on classifying short sentences than the RNN models, but worse in longer sentences. When trained on a small dataset LSTM and GRU had a better result then Naive Bayes. The best performing model was Naive Bayes, which had the highest accuracy score in two out of the three datasets.

In this paper, Natural Language Processing and classification algorithms were used to create a program that automatically can tag different errands that are connected to Fortnox (an IT company based in Växjö) support service. Controlled experiments were conducted to find the best classification algorithm together with different Bag-of-Word pre-processing algorithms to find what was best suited for this problem. All data were provided by Fortnox and were manually labeled with tags connected to it as training and test data. The result of the final algorithm was 69.15% correctly/accurately predicted errands using all original data. When looking at the data that were incorrectly predicted a pattern was noticed where many errands have identical text attached to them. By removing the majority of these errands, the result was increased to 94.08%

The question “who infected whom” is a perennial one in the study of infectious disease dynamics. To understand characteristics of infectious diseases such as how many people will one case produce over the course of infection (the reproductive number), how much time between the infection of two connected cases (the generation interval), and what factors are associated with transmission, one must ascertain who infected whom. The current best practices for linking cases are contact investigations and pathogen whole genome sequencing (WGS). However, these data sources cannot perfectly link cases, are expensive to obtain, and are often not available for all cases in a study. This lack of discriminatory data limits the use of established methods in many existing infectious disease datasets. We developed a method to estimate the relative probability of direct transmission between any two infectious disease cases. We used a subset of cases that have pathogen WGS or contact investigation data to train a model and then used demographic, spatial, clinical, and temporal data to predict the relative transmission probabilities for all case-pairs using a simple machine learning algorithm called naive Bayes. We adapted existing methods to estimate the reproductive number and generation interval to use these probabilities. Finally, we explored the associations between various covariates and transmission and how they related to the associations between covariates and pathogen genetic relatedness. We applied these methods to a tuberculosis outbreak in Hamburg, Germany and to surveillance data in Massachusetts, USA. Through simulations we found that our estimated transmission probabilities accurately classified pairs as links and nonlinks and were able to accurately estimate the reproductive number and the generation interval. We also found that the association between covariates and genetic relatedness captures the direction but not absolute magnitude of the association between covariates and transmission, but the bias was improved by using effect estimates from the naive Bayes algorithm. The methods developed in this dissertation can be used to explore transmission dynamics and estimate infectious disease parameters in established datasets where this was not previously feasible because of a lack of highly discriminatory information, and therefore expand our understanding of many infectious diseases.