Disertaciniame darbe nagrinėjama erdvės – laiko duomenų modeliavimo problema. Dažnai erdvinių duomenų rinkiniai yra gana nedideli, o taškai, kuriuose pasklidę stebėjimai, išsidėstę netaisyklingai. Sprendžiant „erdvinį“ uždavinį, paprastai siekiama inerpoliuoti arba įvertinti erdvinį vidurkį. Laiko eilučių duomenys dažniausiai naudojami ateities reikšmėms prognozuoti. Tuo tarpu erdvės – laiko uždaviniai jungia abu uždavinių tipus. Pasiūlyta keletas originalių erdvinių laiko eilučių modeliavimo metodų. Siūlomi metodai pirmiausia analizuoja vienmates laiko eilutes, o pašalinus laikinę priklausomybė jose, laiko eilučių liekanoms vertinama erdvinė priklausomybė. Tikslas – sudaryti modelį, leidžiantį prognozuoti požymio reikšmę naujame, nestebėtame taške, nauju laiko momentu. Tokio modelio sudarymas remiasi laiko eilučių parametrų erdviniu interpoliavimu. / Space – time data modeling problem is analysed. Often spatial data sets are relatively small, and the points, where observations are taken, are located irregularly. When solving spatial task, usually we are interpolating or estimating the spatial average. Time series data usually are used to predict future values. Meanwhile, the space - time tasks combines both types of tasks. Few original modeling methods of spatial time series are proposed. The proposed methods firstly analyzes the univariate time series, and after removing temporal dependence, spatial dependence in the time series of residuals is measured. Aim of this dissertational work - to create time series model at new unobserved location by incorporating spatial interaction thru spatial interpolation of estimated time series parameters. Such a model is based on the spatial interpolation of time series parameters.

The construction industry has consistently suffered the highest number of fatalities among all human involved industries over the years. Safety managers struggle to prevent injuries and fatalities by monitoring at-risk behavior exhibited by workers and equipment operators. Current methods of identifying and reporting potential hazards on site involve periodic manual inspection, which depends upon personal judgment, is prone to human error, and consumes enormous time and resources. This research presents a framework for automatic identification and analysis of potential hazards by analyzing spatio-temporal data from construction resources. The scope of the research is limited to human-equipment interactions in outdoor construction sites involving ground workers and heavy equipment. A grid-based mapping technique is developed to quantify and visualize potentially hazardous regions caused by resource interactions on a construction site. The framework is also implemented to identify resources that are exposed to potential risk based on their interaction with other resources. Cases of proximity and blind spots are considered in order to create a weight-based scoring approach for mapping hazards on site. The framework is extended to perform ``what-if'' safety analysis for operation planning by iterating through multiple resource configurations. The feasibility of using both real and simulated data is explored. A sophisticated data management and operation analysis platform and a cell-based simulation engine are developed to support the process. This framework can be utilized to improve on-site safety awareness, revise construction site layout plans, and evaluate the need for warning or training workers and equipment operators. It can also be used as an education and training tool to assist safety managers in making better, more effective, and safer decisions.

As cities grow, efficient public transport systems are becoming increasingly important. To offer a more efficient service, public transport providers use systems that predict arrival times of buses, trains and similar vehicles, and present this information to the general public. The accuracy and reliability of these predictions are paramount, since many people depend on them, and erroneous predictions reflect badly on the public transport provider. When public transport vehicles move throughout the cities, they create motion patterns, which describe how their positions change over time. This thesis proposes a way of modeling their motion patterns using Gaussian processes, and investigates whether it is possible to predict the arrival times of public transport buses in Linköping based on their motion patterns. The results are evaluated by comparing the accuracy of the model with a simple baseline model and a recurrent neural network (RNN), and the results show that the proposed model achieves superior performance to that of an RNN trained on the same amounts of data, with excellent explainability and quantifiable uncertainty. However, an RNN is capable of training on much more data than the proposed model in the same amount of time, so in a scenario with large amounts of data the RNN outperforms the proposed model.

Congestion is a major problem in most metropolitan areas and given the increasingrate of urbanization it is likely to be an even more serious problem in the rapidlyexpanding mega cities. One possible method to combat congestion is to provide in-telligent traffic management systems that can in a timely manner inform drivers aboutcurrent or predicted traffic congestions that are relevant to them on their journeys. Thedetection of traffic congestion and the determination of whom to send in advance no-tifications about the detected congestions is the objective of the present research. Byadopting a grid based discretization of space, the proposed system extracts and main-tains traffic flow statistics and mobility statistics from the grid based recent trajectoriesof moving objects, and captures periodical spatio-temporal changes in the traffic flowsand movements by managing statistics for relevant temporal domain projections, i.e.,hour-of-day and day-of-week. Then, the proposed system identifies a directional con-gestion as a cell and its immediate neighbor, where the speed and flow of the objectsthat have moved from the neighbor to the cell significantly deviates from the histori-cal speed and flow statistics. Subsequently, based on one of two notification criteria,namely, Mobility Statistic Criterion (MSC) and Linear Movement Criterion (LMC),the system decides which objects are likely to be affected by the identified conges-tions and sends out notifications to the corresponding objects such that the numberof false negative (missed) and false positive (unnecessary) notifications is minimized.The thesis discusses the design and DBMS-based implementation of the proposedsystem. Empirical evaluations on realistically simulated trajectory data assess the ac-curacy of the methods and test the scalability of the system for varying input sizes andparameter settings. The accuracy assessment results show that the MSC based systemachieves an optimal performance with a true positive notification rate of 0.67 and afalse positive notification rate of 0.05 when min prob equals to 0.35, which is superiorto the performance of the LMC based system. The execution time of- and the spaceused by the system scales linearly with the input size (number of concurrently movingvehicles) and the methods mutually dependent parameters (grid resolution r and RTlength l) that jointly define a spatio-temporal resolution. Within the area of a large  city (40km by 40km), assuming a 60km/h average vehicle speed, the system, runningon a commodity personal computer, can manage the described congestion detectionand three-minute-ahead notification tasks within real-time requirements for 2000 and20000 concurrently moving vehicles for spatio-temporal resolutions (r=100m, l=19)and (r=2km, l=3), respectively.

Finite mixture models have had a profound impact on the history of statistics, contributing to modelling heterogeneous populations, generalizing distributional assumptions, and lately, presenting a convenient framework for classification and clustering. A novel approach, via Gaussian mixture distribution, is introduced for modelling receiver operating characteristic curves. The absence of a closed-form for a functional form leads to employing the Monte Carlo method. This approach performs excellently compared to the existing methods when applied to real data. In practice, the data are often non-normal, atypical, or skewed. It is apparent that non-Gaussian distributions be introduced in order to better fit these data. Two non-Gaussian mixtures, i.e., t distribution and skew t distribution, are proposed and applied to real data. A novel mixture is presented to cluster spatial and temporal data. The proposed model defines each mixture component as a mixture of autoregressive polynomial with logistic links. The new model performs significantly better compared to the most well known model-based clustering techniques when applied to real data. / Thesis / Doctor of Philosophy (PhD)

Continuous monitoring of coastal ecosystems aids in better understanding of their dynamics and inherent harmful effects. As many of these ecosystems prevail over space and time, there is a need for mining this spatio-temporal information for building accurate monitoring and forecast systems. Harmful Algal Blooms (HABs) pose an enormous threat to the U.S. marine habitation and economy in the coastal waters. Federal and state coastal administrators have been devising a state-of-the-art monitoring and forecasting systems for these HAB events. The efficacy of a monitoring and forecasting system relies on the performance of HAB detection. A Machine Learning based Spatio-Temporal data mining approach for the detection of HAB (STML-HAB) events in the region of Gulf of Mexico is proposed in this work. The spatio-temporal cubical neighborhood around the training sample is considered to retrieve relevant spectral information pertaining to both HAB and Non-HAB classes. A unique relevant feature subset combination is derived through evolutionary computation technique towards better classification of HAB from Non-HAB. Kernel based feature transformation and classification is used in developing the model. STML-HAB model gave significant performance improvements over the current optical detection based techniques by highly reducing the false alarm rate with an accuracy of 0.9642 on SeaWiFS data. The developed model is used for prediction on new datasets for further spatio-temporal analyses such as the seasonal variations of HAB, and sequential occurrence of algal blooms. New variability visualizations are introduced to illustrate the dynamic behavior and seasonal variations of HABs from large spatiotemporal datasets. The results outperformed the ensemble of the currently available empirical methods for HAB detection. The ensemble method is implemented by a new approach for combining the empirical models using a probabilistic neural network model. The model is also compared with the results obtained using various feature extraction techniques, spatial neighborhoods and classifiers.

Visual analytics of large amounts of spatio-temporal data is challenging due to the overlap and clutter from movements of multiple objects. A common approach for analyzing such data is to consider how groups of items cluster and move together in space and time. However, most methods for showing Spatio-temporal Cluster (STC) properties, concentrate on a few dimensions of the cluster (e.g. the cluster movement direction or cluster density) and many other properties are not represented. Furthermore, while representing multiple attributes of clusters in a single view existing methods fail to preserve the original shape of the cluster or distort the actual spatial covering of the dataset. In this thesis, I propose a simple yet effective visualization, FlockViz, for showing multiple STC data dimensions in a single view by preserving the original cluster shape. To evaluate this method I develop a framework for categorizing the wide range of tasks involved in analyzing STCs. I conclude this work through a controlled user study comparing the performance of FlockViz with alternative visualization techniques that aid with cluster-based analytic tasks. Finally the exploration capability of FlockViz is demonstrated in some real life data sets such as fish movement, caribou movement, eagle migration, and hurricane movement. The results of the user studies and use cases confirm the advantage and novelty of the novel FlockViz design for visual analytic tasks.

O tamanho dos conjuntos de dados se tornou um grande problema atualmente. À medida que o sensoriamento urbano ganha popularidade, os conjuntos de dados de natureza espacial e temporal se tornam ubíquos, e levantam uma série de questões relacionadas ao armazenamento e gerenciamento destes. Isso também cria uma mudança no paradigma de análise, uma vez que os conjuntos de dados que antes representavam uma única série de medições ordenadas no tempo, agora são compostos por centenas dessas séries, com uma taxa de amostragem que está aumentando constantemente. Além disso, uma vez que os dados urbanos normalmente apresentam disposição geográfica inerente, a maioria das das tarefas requerem o suporte de representações espaciais apropriadas. Este se torna outro problema, visto que as tecnologias de exibição de imagens não avançam na mesma velocidade das tecnologias de sensoriamento, de modo que consequentemente acaba-se tendo mais dados do que espaço visual para representa-los. Após conduzir uma pesquisa exaustiva a respeito de análise de dados temporais e visualização, nós melhoramos uma visualização compacta de series temporais para auxiliar a exploração de grandes conjuntos de dados espaçotemporais. Nossa proposta aproveita a compacticidade de tal representação para permitir o uso de um mapa para representar os atributos espaciais dos dados, de modo coordenado, enquanto representação, de forma compreensível, centenas de series simultaneamente, com total contexto temporal. Nós apresentamos nossa proposta como sendo capaz de auxiliar várias tarefas de caráter exploratório de forma intuitiva. Para defender essa afirmação, nós mostramos como essa ideia foi desenvolvida e melhorada ao longo do desenvolvimento de dois estudos de design visual em diferentes domínios de aplicação, e validamos com a implementação de protótipos que foram usados na análise exploratória de vários conjuntos de dados com 3 representações diferentes. Palavras- / The size of datasets became the major problem in data analysis today. As urban sensing becomes popular, datasets of spatial and temporal nature become ubiquitous, leading to several concerns regarding storage and management. It also creates a shift of paradigm in data analysis, as datasets that once represented a single series of measurements ordered in time are now composed of hundreds of series with ever increasing sampling rates. Also, as urban data usually presents inherent geographic disposition, most analysis tasks requires the support of proper spatial views. It becomes another problem, once that displaying technologies do not advance at the same of pace that sensing technologies do, and consequently, there is usually more data than visual space to represent it. After conducting exhaustive research on temporal data analysis and visualization, we improved a compact visual representation of time series to support the exploration of large spatio-temporal datasets. Our proposal exploits the compactness of such representation to allow the use of a map to represent the spatial properties of the data in a coordinate scheme while presenting, in a comprehensible manner, hundreds of series simultaneously, with full temporal context. We argue that such solution can effectively support many exploratory tasks in an intuitive manner. To support this claim, we show how the idea was conceived, and improved along the development of two design studies from different application domains, and validated by the implementation of prototypes used in the exploratory analysis of several datasets with 3 different data structures.

O tamanho dos conjuntos de dados se tornou um grande problema atualmente. À medida que o sensoriamento urbano ganha popularidade, os conjuntos de dados de natureza espacial e temporal se tornam ubíquos, e levantam uma série de questões relacionadas ao armazenamento e gerenciamento destes. Isso também cria uma mudança no paradigma de análise, uma vez que os conjuntos de dados que antes representavam uma única série de medições ordenadas no tempo, agora são compostos por centenas dessas séries, com uma taxa de amostragem que está aumentando constantemente. Além disso, uma vez que os dados urbanos normalmente apresentam disposição geográfica inerente, a maioria das das tarefas requerem o suporte de representações espaciais apropriadas. Este se torna outro problema, visto que as tecnologias de exibição de imagens não avançam na mesma velocidade das tecnologias de sensoriamento, de modo que consequentemente acaba-se tendo mais dados do que espaço visual para representa-los. Após conduzir uma pesquisa exaustiva a respeito de análise de dados temporais e visualização, nós melhoramos uma visualização compacta de series temporais para auxiliar a exploração de grandes conjuntos de dados espaçotemporais. Nossa proposta aproveita a compacticidade de tal representação para permitir o uso de um mapa para representar os atributos espaciais dos dados, de modo coordenado, enquanto representação, de forma compreensível, centenas de series simultaneamente, com total contexto temporal. Nós apresentamos nossa proposta como sendo capaz de auxiliar várias tarefas de caráter exploratório de forma intuitiva. Para defender essa afirmação, nós mostramos como essa ideia foi desenvolvida e melhorada ao longo do desenvolvimento de dois estudos de design visual em diferentes domínios de aplicação, e validamos com a implementação de protótipos que foram usados na análise exploratória de vários conjuntos de dados com 3 representações diferentes. Palavras- / The size of datasets became the major problem in data analysis today. As urban sensing becomes popular, datasets of spatial and temporal nature become ubiquitous, leading to several concerns regarding storage and management. It also creates a shift of paradigm in data analysis, as datasets that once represented a single series of measurements ordered in time are now composed of hundreds of series with ever increasing sampling rates. Also, as urban data usually presents inherent geographic disposition, most analysis tasks requires the support of proper spatial views. It becomes another problem, once that displaying technologies do not advance at the same of pace that sensing technologies do, and consequently, there is usually more data than visual space to represent it. After conducting exhaustive research on temporal data analysis and visualization, we improved a compact visual representation of time series to support the exploration of large spatio-temporal datasets. Our proposal exploits the compactness of such representation to allow the use of a map to represent the spatial properties of the data in a coordinate scheme while presenting, in a comprehensible manner, hundreds of series simultaneously, with full temporal context. We argue that such solution can effectively support many exploratory tasks in an intuitive manner. To support this claim, we show how the idea was conceived, and improved along the development of two design studies from different application domains, and validated by the implementation of prototypes used in the exploratory analysis of several datasets with 3 different data structures.

Cette thèse a été conduite dans le cadre d’une convention industrielle avec SNCF Réseau. Elle se situe dans le domaine de la géovisualisation d’informations spatio-temporelles, et porte plus particulièrement sur l’élaboration de méthodes de visualisation cartographiques adaptées à l’analyse des impacts des inondations sur le système ferroviaire. Les événements historiques sont une source d’informations importante pour la compréhension et la gestion des risques naturels. La cartographie s’est imposée comme un outil clé pour appréhender les risques dans leurs contextes territoriaux. Mais l’élaboration de visualisations cartographiques qui soient lisibles tout en restituant la complexité des processus survenus lors d’une catastrophe naturelle n’est pas aisée. Les verrous principaux sont la nécessité de représenter à la fois les dimensions spatiales et temporelles des événements issus de risques, le besoin de visualiser les effets dominos, qui conduisent à amplifier les dommages, et la volonté d’adapter les représentations aux besoins et aux capacités cognitives des utilisateurs. Ce travail a conduit à quatre contributions. La première contribution a consisté à formaliser les récits d’inondations dans une ontologie de domaine, qui décrit à la fois les événements issus d’inondations, les effets dominos et leurs impacts sur le système ferroviaire et les mesures de réaction pour ramener le système ferroviaire à l’état d’équilibre. Cinq cas d’inondations historiques ont été instanciés dans le modèle. Leur étude a permis de formuler des principes sémiologiques génériques pour cartographier les récits d’inondations, ce qui constitue la deuxième contribution. La troisième contribution est la production d’une interface de géovisualisation intégrant des représentations graphiques innovantes pour visualiser les temporalités associées aux événements. Cette interface de géovisualisation a fait l’objet d’une expérimentation auprès des experts ferroviaires. Les résultats ont validé les propositions relatives à la représentation du temps mais ont infirmé les propositions de visualisation des effets dominos. Suite à l’analyse des résultats, la quatrième contribution consiste en un modèle de protocole expérimental réutilisable, adapté au test d’interfaces de géovisualisation. / This research was led in an industrial partnership with SNCF Réseau. In the field of geovisualization of spatio-temporal information, it focuses on developing cartographical visualization methods adapted to the analysis of the impacts of floods on the railway system.Historical events are of great help to understand and manage natural risks. Cartography became a key tool to analyze risks in their territorial contexts. But making maps which remain legible while showing all the complexity of risk processes that occurred during natural disaster is not easy. The main challenges are the need to represent both the temporal and the spatial dimensions of risk events, the need to visualize domino-effect, because they often lead to worsen damages, and the will to adapt representations to the cognitive capacities of users.This research resulted in four contributions. The first one is the formalization of flood narratives in a domain ontology, which describes flood events, domino-effects, their impacts on the railway system and also response measures to restore the system. Five case studies of historical floods were instantiated in the model. Resulting from their study, the second contribution consists in generic semiology principles to visualize the narratives of floods on maps. The third contribution is a geovisualization interface, which includes original graphical representations to visualize the temporal features associated with flood events. This geovisualization interface was tested in an experiment with expert users of the railway field. Results confirmed the proposals of representation of time but disconfirm proposals of visualization of domino effects. The analysis of results led to the fourth contribution, which consists in a reusable model of an experimental procedure that is adapted to test geovisualization interfaces.