Machine learning solutions have been successfully used to solve many simple and complex problems. However, their development process still relies on human experts to perform tasks such as data preprocessing, feature engineering and model selection. As the complexity of these tasks increases, so does the demand for automated solutions, namely Automated Machine Learning (AutoML). Most algorithms employed in these systems have hyperparameters whose configuration may directly affect their predictive performance. Therefore, hyperparameter tuning is a recurring task in AutoML systems. This thesis investigated how to efficiently automate hyperparameter tuning by means of Meta-learning. To this end, large-scale experiments were performed tuning the hyperparameters of different classification algorithms, and an enhanced experimental methodology was adopted throughout the thesis to explore and learn the hyperparameter profiles for different classification algorithms. The results also showed that in many cases the default hyperparameter settings induced models that are on par with those obtained by tuning. Hence, a new Meta-learning recommender system was proposed to identify when it is better to use default values and when to tune classification algorithms for each new dataset. The proposed system is capable of generalizing several learning processes into a single modular framework, along with the possibility of assigning different algorithms. Furthermore, a descriptive analysis of model predictions is used to identify which data characteristics affect the necessity for tuning in each one of the algorithms investigated in the thesis. Experimental results also demonstrated that the proposed recommender system reduced the time spent on optimization processes, without reducing the predictive performance of the induced models. Depending on the target algorithm, the Meta-learning recommender system can statistically outperform the baselines. The significance of these results opens a number of new avenues for future work. / Soluções de aprendizado de máquina tem sido cada vez mais usadas com sucesso para resolver problemas dos mais simples aos complexos. Entretanto, o processo de desenvolvimento de tais soluções ainda é um processo que depende da ação de especialistas humanos em tarefas como: pré-processamento dos dados, engenharia de features e seleção de modelos. Consequentemente, quando a complexidade destas tarefas atinge um nível muito alto, há a necessidade de soluções automatizadas, denominadas por Aprendizado de Máquina automatizado (AutoML). A maioria dos algoritmos usados em tais sistemas possuem hiper-parâmetros cujos valores podem afetar diretamente o desempenho preditivo dos modelos gerados. Assim sendo, o ajuste de hiper-parâmetros é uma tarefa recorrente no desenvolvimento de sistems de AutoML. Nesta tese investigou-se a automatização do ajuste de hiper-parâmetros por meio de Meta-aprendizado. Seguindo essa linha, experimentos massivos foram realizados para ajustar os hiper-parâmetros de diferentes algoritmos de classificação. Além disso, uma metodologia experimental aprimorada e adotada ao lngo da tese perimtiu identificar diferentes perfis de ajuste para diferentes algoritmos de classificação. Entretanto, os resultados também mostraram que em muitos casos as configurações default destes algoritmos induziram modelos mais precisos do que os obtidos por meio de ajuste. Assim, foi proposto um novo sistema de recomendação baseado em Meta-learning para identificar quando é melhor realizar o ajuste de parâmetros para os algoritmos de classificação ou apenas usar os valores default. O sistema proposto é capaz de generalizar várias etapas do aprendizado em um único framework modular, juntamente com a possibilidade de avaliar diferentes algoritmos de aprendizado de máquina. As análises descritivas das predições obtidas pelo sistema indicaram quais características podem ser responsáveis por determinar quando o ajuste se faz necessário para cada um dos algoritmos investigados na tese. Os resultados também demonstraram que o sistema recomendador proposto reduziu o tempo gasto com a otimização mantendo o desempenho preditivo dos modelos gerados. Além disso, dependendo do algoritmo de classificação modelado, o sistema foi estatisticamente superior aos baselines. A significância desdes resultados abre um novo número de oportunidades para trabalhos futuros.

Machine learning solutions have been successfully used to solve many simple and complex problems. However, their development process still relies on human experts to perform tasks such as data preprocessing, feature engineering and model selection. As the complexity of these tasks increases, so does the demand for automated solutions, namely Automated Machine Learning (AutoML). Most algorithms employed in these systems have hyperparameters whose configuration may directly affect their predictive performance. Therefore, hyperparameter tuning is a recurring task in AutoML systems. This thesis investigated how to efficiently automate hyperparameter tuning by means of Meta-learning. To this end, large-scale experiments were performed tuning the hyperparameters of different classification algorithms, and an enhanced experimental methodology was adopted throughout the thesis to explore and learn the hyperparameter profiles for different classification algorithms. The results also showed that in many cases the default hyperparameter settings induced models that are on par with those obtained by tuning. Hence, a new Meta-learning recommender system was proposed to identify when it is better to use default values and when to tune classification algorithms for each new dataset. The proposed system is capable of generalizing several learning processes into a single modular framework, along with the possibility of assigning different algorithms. Furthermore, a descriptive analysis of model predictions is used to identify which data characteristics affect the necessity for tuning in each one of the algorithms investigated in the thesis. Experimental results also demonstrated that the proposed recommender system reduced the time spent on optimization processes, without reducing the predictive performance of the induced models. Depending on the target algorithm, the Meta-learning recommender system can statistically outperform the baselines. The significance of these results opens a number of new avenues for future work. / Soluções de aprendizado de máquina tem sido cada vez mais usadas com sucesso para resolver problemas dos mais simples aos complexos. Entretanto, o processo de desenvolvimento de tais soluções ainda é um processo que depende da ação de especialistas humanos em tarefas como: pré-processamento dos dados, engenharia de features e seleção de modelos. Consequentemente, quando a complexidade destas tarefas atinge um nível muito alto, há a necessidade de soluções automatizadas, denominadas por Aprendizado de Máquina automatizado (AutoML). A maioria dos algoritmos usados em tais sistemas possuem hiper-parâmetros cujos valores podem afetar diretamente o desempenho preditivo dos modelos gerados. Assim sendo, o ajuste de hiper-parâmetros é uma tarefa recorrente no desenvolvimento de sistems de AutoML. Nesta tese investigou-se a automatização do ajuste de hiper-parâmetros por meio de Meta-aprendizado. Seguindo essa linha, experimentos massivos foram realizados para ajustar os hiper-parâmetros de diferentes algoritmos de classificação. Além disso, uma metodologia experimental aprimorada e adotada ao lngo da tese perimtiu identificar diferentes perfis de ajuste para diferentes algoritmos de classificação. Entretanto, os resultados também mostraram que em muitos casos as configurações default destes algoritmos induziram modelos mais precisos do que os obtidos por meio de ajuste. Assim, foi proposto um novo sistema de recomendação baseado em Meta-learning para identificar quando é melhor realizar o ajuste de parâmetros para os algoritmos de classificação ou apenas usar os valores default. O sistema proposto é capaz de generalizar várias etapas do aprendizado em um único framework modular, juntamente com a possibilidade de avaliar diferentes algoritmos de aprendizado de máquina. As análises descritivas das predições obtidas pelo sistema indicaram quais características podem ser responsáveis por determinar quando o ajuste se faz necessário para cada um dos algoritmos investigados na tese. Os resultados também demonstraram que o sistema recomendador proposto reduziu o tempo gasto com a otimização mantendo o desempenho preditivo dos modelos gerados. Além disso, dependendo do algoritmo de classificação modelado, o sistema foi estatisticamente superior aos baselines. A significância desdes resultados abre um novo número de oportunidades para trabalhos futuros.

Social Media birgt neben den vielen nutzenbringenden Anwendungsfeldern auch eine Reihe von Gefahren: Der ungefilterte und vor allem unreflektierte Umgang mit einer Vielzahl an Informationsquellen führt zu Phänomenen wie Information Overload oder Cybermobbing, die schließlich in realen gesundheitsgefährdenden Störungen resultieren können. Dabei unterscheiden wir zwischen Störungen der Partizipation und Störungen durch die Exposition und untergliedern diese gemäß pathologischer und sozialwissenschaftlicher Ordnungssysteme. Ebenso wie bei der Erforschung neuer Krankheitsbilder werden hier zunächst eine einheitliche Sprache und eine Taxonomie benötigt, mit der die gefundenen Krankheitsbilder, die Pathologien, korrekt beschrieben und eingeordnet werden können. Der Beitrag liefert hierfür einen ersten Ansatz und schafft damit Voraussetzungen zur Entwicklung informationstechnischer Präventionsmaßnahmen.

Social Media birgt neben den vielen nutzenbringenden Anwendungsfeldern auch eine Reihe von Gefahren: Der ungefilterte und vor allem unreflektierte Umgang mit einer Vielzahl an Informationsquellen führt zu Phänomenen wie Information Overload oder Cybermobbing, die schließlich in realen gesundheitsgefährdenden Störungen resultieren können. Dabei unterscheiden wir zwischen Störungen der Partizipation und Störungen durch die Exposition und untergliedern diese gemäß pathologischer und sozialwissenschaftlicher Ordnungssysteme. Ebenso wie bei der Erforschung neuer Krankheitsbilder werden hier zunächst eine einheitliche Sprache und eine Taxonomie benötigt, mit der die gefundenen Krankheitsbilder, die Pathologien, korrekt beschrieben und eingeordnet werden können. Der Beitrag liefert hierfür einen ersten Ansatz und schafft damit Voraussetzungen zur Entwicklung informationstechnischer Präventionsmaßnahmen.

When a devastating event such as a forest fire occurs, multiple actions have to be taken. The first priority is to ensure people's safety during the fire, then the fire has to be kept under control and finally extinguished. After all of this, what remains is a damaged area in the forest. The objective of this thesis is to evaluate medium and high-resolution satellite imagery for the classification of different burn severities in a wildfire damaged forest. The classification can then be used to plan where to focus restoration efforts after the fire to achieve a safe and economically beneficial usage of the affected area. Trängslet fire in Dalarna and Lillhärdal fire in Härjedalen, the two of the 2018 forest fire sites in Sweden were chosen for this study. Satellite imagery over both study areas at medium spatial resolution from Sentinel-2 were acquired pre-fire in early July, 2018 and post-fire on October 2, 2018 while imagery at high spatial resolution from Pleiades were acquired on September 13, 2018. Image processing, analysis and classification were performed using Google Earth Engine (GEE) and PCI Geomatica. To ensure the quality of the classifications, field data were collected during a field trip to the Lillhärdal area using Open Data Kit (ODK). ODK was used since it is an application that can collect/store georeferenced information and images. The result that this thesis found is that while both the medium and high-resolution classifications achieved accurate results, the Sentinel-2 classification is the most suited method in most cases since it is an easy and automated classification using differential Normalized Burn Ratio (dNBR) compared to the Pleiades classification where a lot of manual work has to be put in. There are however cases where the Pleiades classification would be preferable, such as when the affected area usually is obscured by clouds and Sentinel-2 thus finds it hard to achieve good images and when a good spatial resolution is required to more easily display the classification with the original image. The most accurate result according to the data collected at the site in Lillhärdal also showed that the Pleiades classification had a precise match of 61.54% and a plausible match of 92.31%. This can be compared to the Sentinel-2 classification that had a precise match of 48.72% and a plausible match of 94.87%. These percentages are based on the visual analysis of collected images at the Lillhärdal site compared to the classifications. This thesis could have been improved if more information regarding the groundwork that had been done after the fire, but before the acquiring of the satellite imagery, were available. The result would also most likely be better if a satellite with better spatial resolution than Sentinel-2 but still with near infrared and short-wave infrared bands would have been used. The reason being that dNBR, which gave a good result, only needs those two bands.