A simplistic view of a dataset is that it is collection of numbers. In fact data are much more than that and all data are collected at a set place and time. Often either the location, or the time, is fixed within the dataset and one or both are disregarded. When the place and time of the collection are incorporated into the analysis, the result is a spatio-temporal model. Spatio-temporal data are the focus of this thesis. The majority of the datasets used are radio tracking studies of animals where the objective is to measure the habitat use. Observations are made over a long period of time and a large area. The largest dataset analysed tracks over a hundred animals, in an area larger than 40 square miles, for multiple years. In this context understanding the spatio-temporal relationships between observations is essential. Even data that do not have an obvious spatial component can benefit from spatio-temporal analysis. For example, the data presented on volatility in the stock market do not have an obvious spatial component. The spatial component is the location in the market, not a physical location. Two different methods for measuring and correcting bias are presented. One method relies on direct modelling of the underlying process being observed. The underlying process is animal movement. A model for animal movement is constructed and used to estimate the missing observations that are thought to be the cause of the bias. The second method does not model the animal movement, but instead relies on a Bayesian Hierarchical Model with some simple assumptions. A long running estimation is used to calculate the most likely result without ever directly estimating the underlying equations. In the second section of the thesis two methods for measuring change from shifts in both spatial and temporal location are presented. The methods, Large Diffeomorphic Deformation Metric Mapping (LDDMM) and Diffeomorphic Demons (DD), were originally developed for anatomical data and are adapted here for nonparametric regression surfaces. These are the first applications of LDDMM and DD outside of computational anatomy.

By improving the precision and accuracy of public health surveillance tools, we can improve cost-efficacy and obtain meaningful information to act upon. In this dissertation, we propose statistical methods for improving public health surveillance research. In Chapter 1, we introduce a pooled testing option for HIV prevalence estimation surveys to increase testing consent rates and subsequently decrease non-response bias. Pooled testing is less certain than individual testing, but, if more people to submit to testing, then it should reduce the potential for non-response bias. In Chapter 2, we illustrate technical issues in the design of neonatal tetanus elimination surveys. We address identifying the target population; using binary classification via lot quality assurance sampling (LQAS); and adjusting the design for the sensitivity of the survey instrument. In Chapter 3, we extend LQAS survey designs for monitoring malnutrition for longitudinal surveillance programs. By combining historical information with data from previous surveys, we detect spikes in malnutrition rates. Using this framework, we detect rises in malnutrition prevalence in longitudinal programs in Kenya and the Sudan. In Chapter 4, we develop a computationally efficient geostatistical disease mapping model that naturally handles model fitting issues due to temporal boundary misalignment by assuming that an underlying continuous risk surface induces spatial correlation between areas. We apply our method to assess socioeconomic trends in breast cancer incidence in Los Angeles between 1990 and 2000. In Chapter 5, we develop a statistical framework for addressing statistical uncertainty associated with denominator interpolation and with temporal misalignment in disease mapping studies. We propose methods for assessing the impact of the uncertainty in these predictions on health effects analyses. Then, we construct a general framework for spatial misalignment in regression.

Pagrindinis šio darbo tikslas yra išnagrinėti judančių objektų indeksavimo duomenų bazėse problemas, siūlomus sprendimus bei palyginti keleto iš jų veiksmingumą. Įvairiais pjūviais buvo lyginami praeities duomenis indeksuojantys R ir iš jo išvesti STR bei TB medžiai. Eksperimentai atlikti naudojant sugeneruotus judančių objektų duomenis. Gauti rezultatai parodė, kad indeksų veiksmingas priklauso nuo tam tikrų sąlygų ir aplinkybių, kuriomis jie naudojami. / Over the past few years, there has been a continuous improvement in the wireless communications and the positioning technologies. As a result, tracking the changing positions of continuously moving objects is becoming increasingly feasible and necessary. Databases that deal with objects that change their location and/or shape over time are called spatio-temporal databases. Traditional database approaches for effective information retrieval cannot be used as the moving objects database is highly dynamic. A need for so called spatio-temporal indexing techniques comes to scene. Mainly, by the problem they are addressed to, indices are divided into two groups: a) indexing the past and b) indexing the current and predicted future positions. Also the have been proposed techniques covering both problems. This work is a survey for well known and used indices. Also there is a performance comparison between several past indexing methods. STR Tree, TB Tree and the predecessor of many indices, the R Tree are compared in various aspects using generated datasets of simulated objects movement.

No description available.

With the growth in the size of datasets, data mining has been an important research topic and is receiving substantial interest from both academia and industry for many years. Especially, spatio-temporal data mining, mining knowledge from large amounts of spatio-temporal data, is a highly demanding field because huge amounts of spatio-temporal data are collected in various applications. Therefore, spatio-temporal data mining requires the development of novel data mining algorithms and computational techniques for a successful analysis of large spatio-temporal databases. In this thesis, a spatio-temporal mining technique is proposed and applied on Turkish meteorological data which has been collected from various weather stations in Turkey. This study also includes an analysis and interpretation of spatio-temporal rules generated for Turkish Meteorological data set. We introduce a second level mining technique which is used to define general trends of the patterns according to the spatial changes. Genarated patterns are investigated under different temporal sets in order to monitor the changes of the events with respect to temporal changes.

The analysis of emotions in humans is a field that has been studied for centuries. Through the last decade, multiple approaches towards automatic emotion recognition have been developed to tackle the task of making this analysis autonomous. More specifically, facial expressions in the form of Action Units have been considered until now the most efficient way to recognize emotions. In recent years, applying machine learning for this task has shown outstanding improvements in the accuracy of the solutions. Through this technique, the features can now be automatically learned from the training data, instead of relying on expert domain knowledge and hand-crafted rules. In this thesis, I present Syna and DeepSyna, two models capable of classifying emotional expressions by using both spatial and temporal features. The experimental results demonstrate the effectiveness of Syna in constrained environments, while there is still room for improvement in both constrained and in-the-wild settings. DeepSyna, while addressing this problem, on the other hand suffers from data scarcity and irrelevant transfer learning, which can be solved by future work. / Mänsklig känsloigenkänning har studerats i århundraden. Det senaste årtiondet har mängder av tillvägagångssätt för automatiska processer studerats, för att möjliggöra autonomi; mer specifikt så har ansiktsuttryck i form av Action Units ansetts vara mest effektiva. Maskininlärning har dock nyligen visat att enorma framsteg är möjliga vad gäller bra lösningar på problemen. Så kallade features kan nu automatiskt läras in från träningsdata, även utan expertkunskap och heuristik. Jag presenterar här Syna och DeepSyna, två modeller för ändamålet som använder både spatiala och temporala features. Experiment demonstrerar Synas effektivitet i vissa begränsade omgivningar, medan mycket lämnas att önska vad gäller generella sådana. DeepSyna löser detta men lider samtidigt av databristproblem och onödig så kallad transfer learning, vilket här lämnas till framtida arbete.

Les données permettant de décrire des phénomènes spatio-temporels sont de plus en plus nombreuses. Ces nouvelles données peuvent alors être éloignées de celles habituellement observées pour l'étude de certains phénomènes. Leur analyse, selon une approche hypothético-déductive telle qu'elle est majoritairement effectuée en statistique et dans les SIG, peut ainsi passer sous silence certaines informations insoupçonnées, mais pertinentes, sur les dynamiques de ces phénomènes spatio-temporels.Il peut alors être intéressant de simplement donner à voir les données, pour observer ce qu'elles ont à montrer, avant de les analyser. Ce principe est celui de l'analyse exploratoire: le procédé est de permettre à un utilisateur d'effectuer une exploration libre des données, au moyen de représentations visuelles, afin de mettre en lumière des structures ou des relations insoupçonnées. Aujourd'hui, l'analyse exploratoire est notamment possible au moyen d'environnements de visualisation, intégrant différentes représentations graphiques et cartographiques interactives.Les environnements de visualisation sont majoritairement développés de manière ad hoc, dans le cadre d'une thématique particulière. Or l'émergence constante de nouvelles données incite à promouvoir des méthodes d'analyse applicables à des phénomènes de différentes natures. En fonction de la problématique dans laquelle s'insèrent ces derniers, les dynamiques sur lesquelles va se focaliser l'analyse diffèrent. Analyser un phénomène météorologique dans un but de prévision implique de s’intéresser aux récurrences cycliques du phénomène. Analyser l'évolution d'une population pour la mise en place de politiques publiques implique d’analyser ce phénomène sur le temps long et selon différentes zones de l’espace.Notre objectif est de proposer une méthode d'analyse exploratoire des phénomènes spatio-temporels et de leurs dynamiques, indépendante du thème traité. Pour cela, nous proposons un environnement de géovisualisation, GrAPHiST (Géovisualisation pour l'Analyse des PHenomenes Spatio-Temporels), permettant l'analyse de différentes dynamiques, selon différentes échelles spatiales et temporelles (linéaires ou cycliques). Développer cet environnement implique de s’interroger sur la modélisation du changement dans l’espace, la nature des dynamiques spatio-temporelles à étudier, et les outils visuels et interactifs permettant de les identifier.Ainsi, les contributions de notre recherche se situent à plusieurs niveaux :- une modélisation générique des phénomènes spatio-temporels, sous la forme de séries événementielles;- de nouvelles méthodes de représentations graphiques et interactives, autorisant la recherche et l'identification des dynamiques spatio-temporelles, notamment: l'introduction de diagrammes temporels interactifs permettant la recherche visuelle de récurrences cycliques dans les données spatio-temporelles; l'utilisation de règles de symbologie permettant la visualisation des relations entre les composantes temporelle et spatiale des phénomènes; de nouvelles méthodes de représentations des agrégats d'événements proches, permettant d'identifier des structures dans leur distribution spatio-temporelle;- la formalisation d’une approche d'analyse exploratoire des dynamiques spatio-temporelles, déclinée en plusieurs scénarios selon l’objectif poursuivi.Nous validons notre approche en l'appliquant à l'analyse de différents jeux de données. L'objectif est de vérifier la possibilité d'identifier des dynamiques, relatives au temps linéaire ou cyclique, au moyen de GrAPHiST, et d'illustrer le caractère générique de l'approche, ainsi que les opportunités d'analyse offertes par l'environnement. / Datasets allowing the description of spatio-temporal phenomena are becoming ever more numerous. These new data can be very different from those usually observed for studying spatio-temporal phenomena. An analysis through a hypothetico-deductive approach, like is mainly done in statistic and GIS domains, can ignore some unsuspected, but relevant, information about the dynamics of these spatio-temporal phenomena.It can be interesting then, to just present the data, to observe what they have to show, before analysing them. This is the principle of the exploratory data analysis: the process is to allow a user to freely explore data, through visual representations, in order to highlight unsuspected structures or relationships. Today, exploratory analysis is possible through visualization environments, which integrate different graphic or cartographic interactive representations.Visualization environments are mainly developed in an ad hoc manner, in the context of a particular thematic field. However, the constant appearance of new data encourages promoting analysis methods, which could be applied to several types of phenomena. According to the domain related to these phenomena, the analysis will be focused on different dynamics. Analysing a meteorological phenomenon, in a forecasting purpose, implies a focus on the cyclic recurrences of the phenomenon. Analysing the increase of a population, for the purpose of deciding public policies, implies an analysis of the phenomenon on a long-term, through different spatial areas.Our objective is to propose a method for the exploratory analysis of spatio-temporal phenomena and their dynamics, which would be independent of the topic. In order to achieve this, we propose a geovisualization environment, GrAPHiST (Géovisualisation pour l'Analyse des PHenomenes Spatio-Temporels; Geovisualization for spatio-temporal phenomena analysis), allowing the analysis of several dynamics, through different spatial and temporal (linear or cyclic) scales. Developing this environment implies to focus on how spatial changes are modelled, on the nature of the spatio-temporal dynamics we have to study, and on the visual and interactive tools, which allow the identification of these dynamics.So, the contributions of our research can be found at several levels:a generic modelling approach of spatio-temporal phenomena, in the form of event series;new graphical and interactive representation methods, which allow the searching and the identification of spatio-temporal dynamics, including: the introduction of interactive temporal diagrams, which allow the visual searching of cyclic recurrences in spatio-temporal data; the use of symbology rules, which allow the visualization of relationships between the spatial and temporal components of phenomena; new methods to represent aggregated closed events, which allow to identify structures in their spatio-temporal distribution;the formalization of an exploratory approach for the spatio-temporal dynamics analysis, divided into several scenarios, according to the purpose of the analysis.We validate our proposition by applying it to the analysis of several datasets. The objective is to verify the possibility to identify dynamics, related to linear or cyclic time, through the use of GrAPHiST, and to illustrate the generic aspect of the approach, as well as the analysis opportunities given by the environment.

The transmission of many infectious diseases is affected by climate variations, particularly for diseases spread by arthropod vectors such as malaria and dengue. Previous epidemiological studies have demonstrated statistically significant associations between infectious disease incidence and climate variations. Such research has highlighted the potential for developing climate-based epidemic early warning systems. To establish how much variation in disease risk can be attributed to climatic conditions, non-climatic confounding factors should also be considered in the model parameterisation to avoid reporting misleading climate-disease associations. This issue is sometimes overlooked in climate related disease studies. Due to the lack of spatial resolution and/or the capability to predict future disease risk (e.g. several months ahead), some previous models are of limited value for public health decision making. This thesis proposes a framework to model spatio-temporal variation in disease risk using both climate and non-climate information. The framework is developed in the context of dengue fever in Brazil. Dengue is currently one of the most important emerging tropical diseases and dengue epidemics impact heavily on Brazilian public health services. A negative binomial generalised linear mixed model (GLMM) is adopted which makes allowances for unobserved confounding factors by including spatially structured and unstructured random effects. The model successfully accounts for the large amount of overdispersion found in disease counts. The parameters in this spatio-temporal Bayesian hierarchical model are estimated using Markov Chain Monte Carlo (MCMC). This allows posterior predictive distributions for disease risk to be derived for each spatial location and time period (month/season). Given decision and epidemic thresholds, probabilistic forecasts can be issued, which are useful for developing epidemic early warning systems. The potential to provide useful early warnings of future increased and geographically specific dengue risk is investigated. The predictive validity of the model is evaluated by fitting the GLMM to data from 2001-2007 and comparing probabilistic predictions to the most recent out-of-sample data in 2008-2009. For a probability decision threshold of 30% and the pre-defined epidemic threshold of 300 cases per 100,000 inhabitants, successful epidemic alerts would have been issued for 94% of the 54 microregions that experienced high dengue incidence rates in South East Brazil, during February - April 2008.

In a world of ever increasing amounts of video data, we are forced to abandon traditional methods of scene interpretation by fully manual means. Under such circumstances, some form of automation is highly desirable but this can be a very open ended issue with high complexity. Dealing with such large amounts of data is a non-trivial task that requires efficient selective extraction of parts of a scene which have the potential to develop a higher semantic meaning, alone, or in combination with others. In particular, the types of video data that are in need of automated analysis tend to be outdoor scenes with high levels of activity generated from either foreground or background. Such dynamic scenes add considerable complexity to the problem since we cannot rely on motion energy alone to detect regions of interest. Furthermore, the behaviour of these regions of motion can differ greatly, while still being highly dependent, both spatially and temporally on the movement of other objects within the scene. Modelling these dependencies, whilst eliminating as much redundancy from the feature extraction process as possible are the challenges addressed by this thesis. In the first half, finding the right mechanism to extract and represent meaningful features from dynamic scenes with no prior knowledge is investigated. Meaningful or salient information is treated as the parts of a scene that stand out or seem unusual or interesting to us. The novelty of the work is that it is able to select salient scales in both space and time in which a particular spatio-temporal volume is considered interesting relative to the rest of the scene. By quantifying the temporal saliency values of regions of motion, it is possible to consider their importance in terms of both the long and short-term. Variations in entropy over spatio-temporal scales are used to select a context dependent measure of the local scene dynamics. A method of quantifying temporal saliency is devised based on the variation of the entropy of the intensity distribution in a spatio-temporal volume over incraeasing scales. Entropy is used over traditional filter methods since the stability or predictability of the intensity distribution over scales of a local spatio-temporal region can be defined more robustly relative to the context of its neighbourhood, even for regions exhibiting high intensity variation due to being extremely textured. Results show that it is possible to extract both locally salient features as well as globally salient temporal features from contrasting scenerios. In the second part of the thesis, focus will shift towards binding these spatio-temporally salient features together so that some semantic meaning can be inferred from their interaction. Interaction in this sense, refers to any form of temporally correlated behaviour between any salient regions of motion in a scene. Feature binding as a mechanism for interactive behaviour understanding is particularly important if we consider that regions of interest may not be treated as particularly significant individually, but represent much more semantically when considered in combination. Temporally correlated behaviour is identified and classified using accumulated co-occurrences of salient features at two levels. Firstly, co-occurrences are accumulated for spatio-temporally proximate salient features to form a local representation. Then, at the next level, the co-occurrence of these locally spatio-temporally bound features are accumulated again in order to discover unusual behaviour in the scene. The novelty of this work is that there are no assumptions made about whether interacting regions should be spatially proximate. Furthermore, no prior knowledge of the scene topology is used. Results show that it is possible to detect unusual interactions between regions of motion, which can visually infer higher levels of semantics. In the final part of the thesis, a more specific investigation of human behaviour is addressed through classification and detection of interactions between 2 human subjects. Here, further modifications are made to the feature extraction process in order to quantify the spatiotemporal saliency of a region of motion. These features are then grouped to find the people in the scene. Then, a loose pose distribution model is extracted for each person for finding salient correlations between poses of two interacting people using canonical correlation analysis. These canonical factors can be formed into trajectories and used for classification. Levenshtein distance is then used to categorise the features. The novelty of the work is that the interactions do not have to be spatially connected or proximate for them to be recognised. Furthermore, the data used is outdoors and cluttered with non-stationary background. Results show that co-occurrence techniques have the potential to provide a more generalised, compact, and meaningful representation of dynamic interactive scene behaviour.

Mining spatio-temporal reachable regions aims to find a set of road segments from massive trajectory data, that are reachable from a user-specified location and within a given temporal period. Accurately extracting such spatio-temporal reachable area is vital in many urban applications, e.g., (i) location-based recommendation, (ii) location-based advertising, and (iii) business coverage analysis. The traditional approach of answering such queries essentially performs a distance-based range query over the given road network, which have two main drawbacks: (i) it only works with the physical travel distances, where the users usually care more about dynamic traveling time, and (ii) it gives the same result regardless of the querying time, where the reachable area could vary significantly with different traffic conditions. Motivated by these observations, in this thesis, we propose a data- driven approach to formulate the problem as mining actual reachable region based on real historical trajectory dataset. The main challenge in our approach is the system efficiency, as verifying the reachability over the massive trajectories involves huge amount of disk I/Os. In this thesis, we develop two indexing structures: 1) spatio-temporal index (ST-Index) and 2) connection index (Con-Index) to reduce redundant trajectory data access operations. We also propose a novel query processing algorithm with: 1) maximum bounding region search, which directly extracts a small searching region from the index structure and 2) trace back search, which refines the search results from the previous step to find the final query result. Moreover, our system can also efficiently answer the spatio-temporal reachability query with multiple query locations by skipping the overlapped area search. We evaluate our system extensively using a large-scale real taxi trajectory data in Shenzhen, China, where results demonstrate that the proposed algorithms can reduce 50%-90% running time over baseline algorithms.