Proximity scaling methods (e.g., multidimensional scaling) represent objects in a low dimensional configuration so that fitted distances between objects optimally approximate multivariate proximities. Next to finding the optimal configuration the goal is often also to assess groups of objects from the configuration. This can be difficult if the optimal configuration lacks clusteredness (coined c-clusteredness). We present Cluster Optimized Proximity Scaling (COPS), which attempts to solve this problem by finding a configuration that exhibts c-clusteredness. In COPS, a flexible scaling loss function (p-stress) is combined with an index that quantifies c-clusteredness in the solution, the OPTICS Cordillera. We present two variants of combining p-stress and Cordillera, one for finding the configuration directly and one for metaparameter selection for p-stress. The first variant is illustrated by scaling Californian counties with respect to climate change related natural hazards. We identify groups of counties with similar risk profiles and find that counties that are in high risk of drought are socially vulnerable. The second variant is illustrated by finding a clustered nonlinear representation of countries according to their history of banking crises from 1800 to 2010. (authors' abstract) / Series: Discussion Paper Series / Center for Empirical Research Methods

M.Ing. (Mechanical) / The goal of this study was to determine whether a SLIRN direct reduction process could be modelled with a neural network. The full name of the SLIRN process is the Stelco, Lurgi, Republic Steel, and National Leadprocess. A parallel goal was to identify, and test an alternative method to reduce the dimensionality of a model. A neural network software package named Process Insights was used to model the process. Two independent data reduction methods were used along with various Process Insights functions, to build, train, and test models. The best model produced by each of the two data reduction methods was used to report on. The results showed that a SLIRN direct reduction process could be modelled successfully with a neural network. The large number of variables normally identified with such a process can be reduced without significant loss in model performance, The results also showed that the removal of the most significant variable does not affect the model accuracy significantly, which bodes well for the fault tolerance of the model in terms of individual sensor failures. The Process Insights functions important to the modelling process were highlighted.

The mortality is historically one of the most important demographic indicator and definitely reflects the maturity of each country. The objective of this diploma thesis is the comparison of mortality rates in analyzed countries around the world over time and among each other using the principle component analysis that allows assessing data different way. The big advantage of this method is minimal loss of information and quite understandable interpretation of mortality in each country. This thesis offers several interesting graphical outputs, that for example confirm higher mortality rate in Eastern European countries compared to Western European countries and show that Czech republic is country where mortality has fallen most in context of post-communist countries between 1990 and 2010. Source of the data is Human Mortality Database and all data were processed in statistical tool SPSS.

In this work we deal with cluster analysis for functional data. Functional data contain a set of subjects that are characterized by repeated measurements of a variable. Based on these measurements we want to split the subjects into groups (clusters). The subjects in a single cluster should be similar and differ from subjects in the other clusters. The first approach we use is the reduction of data dimension followed by the clustering method K-means. The second approach is to use a finite mixture of normal linear mixed models. We estimate parameters of the model by maximum likelihood using the EM algorithm. Throughout the work we apply all described procedures to real meteorological data.

La technologie des réseaux de capteurs sans fil démontre qu'elle peut être très utile dans de nombreuses applications. Ainsi chaque jour voit émerger de nouvelles réalisations dans la surveillance de notre environnement comme la détection des feux de forêt, l'approvisionnement en eau. Les champs d'applications couvrent aussi des domaines émergents et sensibles pour la population avec les soins aux personnes âgées ou les patients récemment opérés dans le cadre. L'indépendance des architectures RCSFs par rapport aux infrastructures existantes permet aux d'être déployées dans presque tous les sites afin de fournir des informations temporelles et spatiales. Dans les déploiements opérationnels le bon fonctionnement de l'architecture des réseaux de capteurs sans fil ne peut être garanti que si certains défis sont surmontés. La minisation de l'énergie consommée en fait partie. La limitation de la durée de vie des nœuds de capteurs est fortement couplée à l'autonomie de la batterie et donc à l'optimisation énergétique des nœuds du réseau. Nous présenterons plusieurs propositions à ces problèmes dans le cadre de cette thèse. En résumé, les contributions qui ont été présentées dans cette thèse, abordent la durée de vie globale du réseau, l'exploitation des messages de données redondantes et corrélées et enfin le fonctionnement nœud lui-même. Les travaux ont conduit à la réalisation d'algorithmes de routage hiérarchiques et de filtrage permettant la suppression des redondances. Ils s'appuient sur les corrélations spatio-temporelles des données mesurées. Enfin, une implémentation de ce réseau de capteurs multi-sauts intégrant ces nouvelles fonctionnalités est proposée. / Wireless sensor networks (WSNs) technology have been demonstrated to be a usefulmeasurement system for numerous bath indoor and outdoor applications. There is avast amount of applications that are operating with WSN technology, such asenvironmental monitoring, for forest fire detection, weather forecasting, water supplies, etc. The independence nature of WSNs from the existing infrastructure. Virtually, the WSNs can be deployed in any sort of location, and provide the sensor samples accordingly in bath time and space. On the contrast, the manual deployments can only be achievable at a high cost-effective nature and involve significant work. ln real-world applications, the operation of wireless sensor networks can only be maintained, if certain challenges are overcome. The lifetime limitation of the distributed sensor nodes is amongst these challenges, in order to achieve the energy optimization. The propositions to the solution of these challenges have been an objective of this thesis. ln summary, the contributions which have been presented in this thesis, address the system lifetime, exploitation of redundant and correlated data messages, and then the sensor node in terms of usability. The considerations have led to the simple data redundancy and correlated algorithms based on hierarchical based clustering, yet efficient to tolerate bath the spatio-temporal redundancies and their correlations. Furthermore, a multihop sensor network for the implementation of propositions with more features, bath the analytical proofs and at the software level, have been proposed.