Littérature scientifique sur le sujet « Spatiotemporal granularitie »

Spatiotemporal data prediction is of great significance in the fields of smart cities and smart manufacturing. Current spatiotemporal data prediction models heavily rely on traditional spatial views or single temporal granularity, which suffer from missing knowledge, including dynamic spatial correlations, periodicity, and mutability. This paper addresses these challenges by proposing a multi-layer attention-based predictive model. The key idea of this paper is to use a multi-layer attention mechanism to model the dynamic spatial correlation of different features. Then, multi-granularity historical features are fused to predict future spatiotemporal data. Experiments on real-world data show that the proposed model outperforms six state-of-the-art benchmark methods.

Studying population prediction under micro-spatiotemporal granularity is of great significance for modern and refined urban traffic management and emergency response to disasters. Existing population studies are mostly based on census and statistical yearbook data due to the limitation of data collecting methods. However, with the advent of techniques in this information age, new emerging data sources with fine granularity and large sample sizes have provided rich materials and unique venues for population research. This article presents a new population prediction model with micro-spatiotemporal granularity based on the long short-term memory (LSTM) and cellular automata (CA) models. We aim at designing a hybrid data-driven model with good adaptability and scalability, which can be used in more refined population prediction. We not only try to integrate these two models, aiming to fully mine the spatiotemporal characteristics, but also propose a method that fuses multi-source geographic data. We tested its functionality using the data from Chongming District, Shanghai, China. The results demonstrated that, among all scenarios, the model trained by three consecutive days (ordinary dates), with the granularity of one hour, incorporated with road networks, achieves the best performance (0.905 as the mean absolute error) and generalization capability.

Electricity is hidden within wires and networks only revealing its quantity and flow when metered. The making of its properties into data is therefore particularly important to the relations that are formed around electricity as a produced and managed phenomenon. We propose approaching all metering as a situated activity, a form of quantification work in which data is made and becomes mobile in particular spatial and temporal terms, enabling its entry into data infrastructures and schemes of evaluation and value production. We interrogate the transition from the pre-digital into the making of bigger, more spatiotemporally granular electricity data, through focusing on those actors selling and materialising new metering technologies, data infrastructures and services for larger businesses and public sector organisations in the UK. We examine the claims of truth and visibility that accompany these shifts and their enrolment into management techniques that serve to more precisely apportion responsibility for, and evaluate the status of, particular patterns and instances of electricity use. We argue that whilst through becoming Big Data electricity flow is now able to be known and given identity in significantly new terms, enabling new relations to be formed with the many heterogeneous entities implicated in making and managing energy demand, it is necessary to sustain some ambivalence as to the performative consequences that follow for energy governance. We consider the wider application of our conceptualisation of metering, reflecting on comparisons with the introduction of new metering systems in domestic settings and as part of other infrastructural networks.

Effective quantification of visitation is important for understanding many impacts of the COVID-19 pandemic on national parks and other protected areas. In this study, we mapped and analyzed the spatiotemporal patterns of visitation for six national parks in the western U.S., taking advantage of large mobility records sampled from mobile devices and released by SafeGraph as part of their Social Distancing Metric dataset. Based on comparisons with visitation statistics released by the U.S. National Park Service, our results confirmed that mobility records from digital devices can effectively capture park visitation patterns but with much finer spatiotemporal granularity. In general, triggers of visitation changes corresponded well with the parks’ management responses to COVID-19, with all six parks showing dramatic decreases in the number of visitors (compared to 2019) beginning in March 2020 and continuing through April and May. As restrictions were eased to promote access to the parks and the benefits associated with outdoor recreation, visitation in 2020 approached or even passed that from 2019 by late summer or early autumn at most of the parks. The results also revealed that parks initially saw the greatest increases in visitation after reopening originating from nearby states, with visitorship coming from a broader range of states as time passed. Our study highlights the capability of mobility data for providing spatiotemporally explicit knowledge of place visitation.

This paper presents a data fusion framework for processing and integrating data collected from heterogeneous sources on motorways to generate short-term predictions. Considering the heterogeneity in spatiotemporal granularity in data from different sources, an adaptive kernel-based smoothing method was first used to project all data onto a common space–time grid. The data were then integrated through a Kalman filter framework build based on the cell transmission model for generating short-term traffic state prediction. The algorithms were applied and tested with real traffic data collected from the California I-880 corridor in the San Francisco Bay Area from the Mobile Century experiment. Results revealed that the proposed fusion algorithm can work with data sources that are different in their spatiotemporal granularity and improve the accuracy of state estimation through incorporating multiple data sources. The present work contributed to the field of traffic engineering and management with the application of big data analytics.

Abstract. Student behavior research can improve learning efficiency, provide decision evidences for infrastructure management. Existing campus-scale behavioral analysis work have not taken into account the students characteristics and spatiotemporal pattern. Moreover, the visualization methods are weak in wholeness, intuitiveness and interactivity perspectives. In this paper, we design a geospatial dashboard-based student behavior analysis and visualization system considering students characteristics and spatiotemporal pattern. This system includes four components: user monitoring, data mining analysis, behavior prediction and spatiotemporal visualization. Furthermore, a deep learning model based on LSTNet to predict student behaviour. Our work takes WiFi log data of a university in Beijing as dataset. The results show that this system can identify student behavior patterns at a finer granularity by visualization method, which is helpful in improving learning and living efficiency.

As a typical spatiotemporal problem, there are three main challenges in traffic forecasting. First, the road network is a nonregular topology, and it is difficult to extract complex spatial dependence accurately. Second, there are short- and long-term dependencies between traffic dates. Third, there are many other factors besides the influence of spatiotemporal dependence, such as semantic characteristics. To address these issues, we propose a spatiotemporal DeepWalk gated recurrent unit model (ST-DWGRU), a deep learning framework that fuses spatial, temporal, and semantic features for traffic speed forecasting. In the framework, the spatial dependency between nodes of an entire road network is extracted by graph convolutional network (GCN), whereas the temporal dependency between speeds is captured by a gated recurrent unit network (GRU). DeepWalk is used to extract semantic information from road networks. Three publicly available datasets with different time granularities of 15, 30, and 60 min are used to validate the short- and long-time prediction effect of this model. The results show that the ST-DWGRU model significantly outperforms the state-of-the-art baselines.

In molti campi di ricerca, i ricercatori hanno la necessità di memorizzare, gestire e interrogare dati spazio-temporali. Tali dati sono classici dati alfanumerici arricchiti però con una o più componenti temporali, spaziali e spazio-temporali che, con diversi possibili significati, li localizzano nel tempo e/o nello spazio. Ambiti in cui tali dati spazio-temporali devono essere raccolti e gestiti sono, per esempio, la gestione del territorio o delle risorse naturali, l'epidemiologia, l'archeologia e la geografia. Più in dettaglio, per esempio nelle ricerche epidemiologiche, i dati spazio-temporali possono servire a rappresentare diversi aspetti delle malattie e delle loro caratteristiche, quali per esempio la loro origine, espansione ed evoluzione e i fattori di rischio potenzialmente connessi alle malattie e al loro sviluppo. Le componenti spazio-temporali dei dati possono essere considerate come dei "meta-dati" che possono essere sfruttati per introdurre nuovi tipi di analisi sui dati stessi. La gestione di questi "meta-dati" può avvenire all'interno di diversi framework proposti in letteratura. Uno dei concetti proposti a tal fine è quello delle granularità. In letteratura c'è ampio consenso sul concetto di granularità temporale, di cui esistono framework basati su diversi approcci. D'altro canto, non esiste invece un consenso generale sulla definizione di un framework completo, come quello delle granularità temporali, per le granularità spaziali e spazio-temporali. Questa tesi ha lo scopo di riempire questo vuoto proponendo un framework per le granularità spaziali e, basandosi su questo e su quello già presente in letteratura per le granularità temporali, un framework per le granularità spazio-temporali. I framework proposti vogliono essere completi, per questo, oltre alle definizioni dei concetti di granularità spaziale e spazio-temporale, includono anche la definizione di diversi concetti legati alle granularità, quali per esempio le relazioni e le operazioni tra granularità. Le relazioni permettono di conoscere come granularità diverse sono legate tra loro, costruendone anche una gerarchia. Tali informazioni sono poi utili al fine di conoscere se e come è possibile confrontare dati associati e rappresentati con granularità diverse. Le operazioni permettono invece di creare nuove granularità a partire da altre granularità già definite nel sistema, manipolando o selezionando alcune loro componenti. Basandosi su questi framework, l'obiettivo della tesi si sposta poi sul mostrare come le granularità possano essere utilizzate per arricchire basi di dati spazio-temporali già esistenti al fine di una loro migliore e più ricca gestione e interrogazione. A tal fine, proponiamo qui una base di dati per la gestione dei dati riguardanti le granularità temporali, spaziali e spazio-temporali. Nella base di dati proposta possono essere rappresentate tutte le componenti di una granularità come definito nei framework proposti. La base di dati può poi essere utilizzata per estendere una base di dati spazio-temporale esistente aggiungendo alle tuple di quest'ultima delle referenze alle granularità dove quei dati possono essere localizzati nel tempo e/o nel spazio. Per dimostrare come ciò possa essere fatto, nella tesi introduciamo la base di dati sviluppata ed utilizzata dal Servizio Psichiatrico Territoriale (SPT) di Verona. Tale base di dati memorizza le informazioni su tutti i pazienti venuti in contatto con l'SPT negli ultimi 30 anni e tutte le informazioni sui loro contatti con il servizio stesso (per esempio: chiamate telefoniche, visite a domicilio, ricoveri). Parte di tali informazioni hanno una componente spazio-temporale e possono essere quindi analizzate studiandone trend e pattern nel tempo e nello spazio. Nella tesi quindi estendiamo questa base di dati psichiatrica collegandola a quella proposta per la gestione delle granularità. A questo punto i dati psichiatrici possono essere interrogati anche sulla base di vincoli spazio-temporali basati su granularità. L'interrogazione di dati spazio-temporali associati a granularità richiede l'utilizzo di un linguaggio d'interrogazione che includa, oltre a strutture, operatori e funzioni spazio-temporali per la gestione delle componenti spazio-temporali dei dati, anche costrutti per l'utilizzo delle granularità nelle interrogazioni. Quindi, partendo da un linguaggio d'interrogazione spazio-temporale già presente in letteratura, in questa tesi proponiamo anche un linguaggio d'interrogazione che permetta ad un utente di recuperare dati da una base di dati spazio-temporale anche sulla base di vincoli basati su granularità. Il linguaggio viene introdotto fornendone la sintassi e la semantica. Inoltre per mostrare l'effettivo ruolo delle granularità nell'interrogazione di una base di dati clinica, mostreremo diversi esempi di interrogazioni, scritte con il linguaggio d'interrogazione proposto, sulla base di dati psichiatrica dell'SPT di Verona. Tali interrogazioni spazio-temporali basate su granularità possono essere utili ai ricercatori ai fini di analisi epidemiologiche dei dati psichiatrici.
In several research fields, temporal, spatial, and spatio-temporal data have to be managed and queried with several purposes. These data are usually composed by classical data enriched with a temporal and/or a spatial qualification. For instance, in epidemiology spatio-temporal data may represent surveillance data, origins of disease and outbreaks, and risk factors. In order to better exploit the time and spatial dimensions, spatio-temporal data could be managed considering their spatio-temporal dimensions as meta-data useful to retrieve information. One way to manage spatio-temporal dimensions is by using spatio-temporal granularities. This dissertation aims to show how this is possible, in particular for epidemiological spatio-temporal data. For this purpose, in this thesis we propose a framework for the definition of spatio-temporal granularities (i.e., partitions of a spatio-temporal dimension) with the aim to improve the management and querying of spatio-temporal data. The framework includes the theoretical definitions of spatial and spatio-temporal granularities (while for temporal granularities we refer to the framework proposed by Bettini et al.) and all related notions useful for their management, e.g., relationships and operations over granularities. Relationships are useful for relating granularities and then knowing how data associated with different granularities can be compared. Operations allow one to create new granularities from already defined ones, manipulating or selecting their components. We show how granularities can be represented in a database and can be used to enrich an existing spatio-temporal database. For this purpose, we conceptually and logically design a relational database for temporal, spatial, and spatio-temporal granularities. The database stores all data about granularities and their related information we defined in the theoretical framework. This database can be used for enriching other spatio-temporal databases with spatio-temporal granularities. We introduce the spatio-temporal psychiatric case register, developed by the Verona Community-based Psychiatric Service (CPS), for storing and managing information about psychiatric patient, their personal information, and their contacts with the CPS occurred in last 30 years. The case register includes both clinical and statistical information about contacts, that are also temporally and spatially qualified. We show how the case register database can be enriched with spatio-temporal granularities both extending its structure and introducing a spatio-temporal query language dealing with spatio-temporal data and spatio-temporal granularities. Thus, we propose a new spatio-temporal query language, by defining its syntax and semantics, that includes ad-hoc features and constructs for dealing with spatio-temporal granularities. Finally, using the proposed query language, we report several examples of spatio-temporal queries on the psychiatric case register showing the ``usage'' of granularities and their role in spatio-temporal queries useful for epidemiological studies.

Crimes, forest fires, accidents, infectious diseases, human interactions with mobile devices (e.g., tweets) are being logged as spatiotemporal events. For each event, its spatial location, time and related attributes are known with high levels of detail (LoDs). The LoD of analysis plays a crucial role in the user’s perception of phenomena. From one LoD to another, some patterns can be easily perceived or different patterns may be detected, thus requiring modeling phenomena at different LoDs as there is no exclusive LoD to study them. Granular computing emerged as a paradigm of knowledge representation and processing, where granules are basic ingredients of information. These can be arranged in a hierarchical alike structure, allowing the same phenomenon to be perceived at different LoDs. This PhD Thesis introduces a formal Theory of Granularities (ToG) in order to have granules defined over any domain and reason over them. This approach is more general than the related literature because these appear as particular cases of the proposed ToG. Based on this theory we propose a granular computing approach to model spatiotemporal phenomena at multiple LoDs, and called it a granularities-based model. This approach stands out from the related literature because it models a phenomenon through statements rather than just using granules to model abstract real-world entities. Furthermore, it formalizes the concept of LoD and follows an automated approach to generalize a phenomenon from one LoD to a coarser one. Present-day practices work on a single LoD driven by the users despite the fact that the identification of the suitable LoDs is a key issue for them. This PhD Thesis presents a framework for SUmmarizIng spatioTemporal Events (SUITE) across multiple LoDs. The SUITE framework makes no assumptions about the phenomenon and the analytical task. A Visual Analytics approach implementing the SUITE framework is presented, which allow users to inspect a phenomenon across multiple LoDs, simultaneously, thus helping to understand in what LoDs the phenomenon perception is different or in what LoDs patterns emerge.

This chapter introduces an object-oriented methodology for Geographical Information Systems (GIS) development. It argues that a COTS-based development methodology combined with the UML, can be extended to support the spatiotemporal peculiarities that characterize GIS applications. The author hopes that by typifying both enterprises and developments, and, with a thorough knowledge of the software component granularity in the GIS domain, it will be possible to extend and adapt the proposed COTS-based methodologies to cover the full lifecycle. Moreover, some recommendations are outlined to translate the methodology to the commercial iCASE Rational Suite Enterprise and its relationships with tool kits proposed by some GIS COTS vendors.