Geoinformatica最新文献

Online businesses and websites have recently become the main target of fake reviews, where fake reviews are intentionally composed to manipulate the business ratings positively or negatively. Most of existing works to detect fake reviews are supervised methods, whose performance highly depends on the amount, quality, and variety of the labeled data, which are often non-trivial to obtain in practice. In this paper, we propose a semi-supervised label sparsity-tolerant framework, LENS, for fake review detection by mining spatial knowledge and learning distributions of embedded topics. LENS builds on two key observations. (1) Spatial knowledge revealed in spatial entities and their co-occurring latent topic distributions may indicate the review authenticity. (2) Distributions of the embedded topics (the contextual distribution) may exhibit important patterns to differentiate between real and fake reviews. Specifically, LENS first extracts embeddings for spatial named entities using a knowledge base trained from Wikipedia webpages. Second, LENS represents each input token as a distribution over the learned latent topics in the embedded topic space. To bypass the differentiation difficulty, LENS builds on two discriminators in the actor-critic architecture using reinforcement learning. Extensive experiments using the real-world spatial and non-spatial datasets show that LENS consistently outperformed the state-of-the-art semi-supervised fake review detection methods on few labels at all different labeling rates for real and fake reviews, respectively, in a label-starving setting.

The majority of real-world processes are spatiotemporal, and the data generated by them exhibits both spatial and temporal evolution. Weather is one of the most essential processes in this domain, and weather forecasting has become a crucial part of our daily routine. Weather data analysis is considered the most complex and challenging task. Although numerical weather prediction models are currently state-of-the-art, they are resource-intensive and time-consuming. Numerous studies have proposed time series-based models as a viable alternative to numerical forecasts. Recent research in the area of time series analysis indicates significant advancements, particularly regarding the use of state-space-based models (white box) and, more recently, the integration of machine learning and deep neural network-based models (black box). The most famous examples of such models are RNNs and transformers. These models have demonstrated remarkable results in the field of time-series analysis and have demonstrated effectiveness in modelling temporal correlations. It is crucial to capture both temporal and spatial correlations for a spatiotemporal process, as the values at nearby locations and time affect the values of a spatiotemporal process at a specific point. This self-contained paper explores various regional data-driven weather forecasting methods, i.e., forecasting over multiple latitude-longitude points (matrix-shaped spatial grid) to capture spatiotemporal correlations. The results showed that spatiotemporal prediction models reduced computational costs while improving accuracy. In particular, the proposed tensor train dynamic mode decomposition-based forecasting model has comparable accuracy to the state-of-the-art models without the need for training. We provide convincing numerical experiments to show that the proposed approach is practical.

Multi-modal trajectory representation learning aims to convert raw trajectories into low-dimensional embeddings to facilitate downstream trajectory analysis tasks. However, existing methods focus on spatio-temporal trajectories and often neglect additional modal features such as textual or imagery data. Moreover, these methods do not fully consider the correlations among different modal features and the relationships among trajectories, thus hindering the generation of generic and semantically enriched representations. To address these limitations, we propose a generic Contrastive Learning-based Multi-modal Trajectory Representation framework, termed CLMTR. Specifically, we incorporate intra- and inter-trajectory contrastive learning components to capture the correlations among diverse modal features and the intricate relationships among trajectories, obtaining generic and semantically enriched trajectory representations. We develop multi-modal feature embedding and attention-based fusion approaches to capture the multi-modal characteristics and adaptively obtain the unified embeddings. Experimental results on two real-world datasets demonstrate the superior performance of CLMTR over state-of-the-art methods in three downstream tasks.

Traffic forecasting is the foundation and core task of Intelligent Transportation Systems (ITS). Due to the powerful ability of Graph Neural Network (GNN) to capture topological features, recently, it is commonly used in traffic forecasting to capture spatial features of road networks. Although existing GNN based traffic forecasting methods have achieved satisfactory results, they are still plagued by the following problems: (1) Traffic time-series usually contains complex periodic features, but they only model 1D time features, ignoring multi-periodic information in traffic data. (2) There are multivariate higher-order correlations among nodes in road networks, but they only preserve the pairwise connections by simple graphs, neglecting the higher-order multivariate correlations. (3) They cannot adaptively capture unique patterns of specific areas, only learn the shared patterns of traffic time-series. To solve the above problems, we propose a Periodicity aware spatial-temporal Adaptive Hypergraph Neural Network (PAHNN). Firstly, a temporal multi-periodic block is designed to capture the 2D-variations of traffic time-series to extract multi-periodic features and complex temporal patterns. Then, we propose a spatial adaptive hypergraph block to model spatial multivariate correlations among nodes via hypergraph neural networks. Adaptive selection of hypergraph networks for different data can extract specific spatial patterns of different traffic areas. Finally, extensive experiments are conducted on two types of forecasting tasks to evaluate the effectiveness and accuracy of our model.

Accurate shared micromobility demand predictions are essential for transportation planning and management. Although deep learning methods provide robust mechanisms to tackle demand forecasting challenges, current models based on graph neural networks suffer from limited scalability and high computational cost. There is both a need and significant potential to enhance the accuracy and efficiency of existing shared micromobility demand forecasting models. To fill these research gaps, this paper proposes a deep learning model named Interactive Convolutional Network (ICN) to forecast spatiotemporal travel demand for shared micromobility. The proposed model develops a novel channel dilation method by utilizing multi-dimensional spatial information (i.e., demographics, functionality, and transportation supply) based on travel behavior knowledge for building the deep learning model. We use the convolution operation to process the dilated tensor to simultaneously capture temporal and spatial dependencies. Based on a binary-tree-structured architecture and interactive convolution, the ICN model extracts features at different temporal resolutions and then generates predictions using a fully-connected layer. We conducted two practical case studies from Chicago, IL, and Austin, TX to test the proposed model. The results show that the ICN model significantly outperforms all benchmark models. The model predictions have the potential to assist micromobility operators in developing efficient vehicle rebalancing strategies, while also providing cities with guidance on enhancing the management of their shared micromobility system.

In recent years, the maritime domain has experienced tremendous growth due to the exploitation of big traffic data. Particular emphasis has been placed on deep learning methodologies for decision-making. Accurate Vessel Traffic Flow Forecasting (VTFF) is essential for optimizing navigation efficiency and proactively managing maritime operations. In this work, we present a distributed Unified Approach for VTFF (dUA-VTFF), which employs Transformer models and leverages the Apache Spark big data distributed processing framework to learn from historical maritime data and predict future traffic flows over a time horizon of up to 30 min. Particularly, dUA-VTFF leverages vessel timestamped locations along with future vessel locations produced by a Vessel Route Forecasting model. These data are arranged into a spatiotemporal grid to formulate the traffic flows. Subsequently, through the Apache Spark, each grid cell is allocated to a computing node, where appropriately designed Transformer-based models forecast traffic flows in a distributed framework. Experimental evaluations conducted on real Automatic Identification System (AIS) datasets demonstrate the improved efficiency of the dUA-VTFF compared to state-of-the-art traffic flow forecasting methods.

Trajectory data combines the complexities of time series, spatial data, and (sometimes irrational) movement behavior. As data availability and computing power have increased, so has the popularity of deep learning from trajectory data. This review paper provides the first comprehensive overview of deep learning approaches for trajectory data. We have identified eight specific mobility use cases which we analyze with regards to the deep learning models and the training data used. Besides a comprehensive quantitative review of the literature since 2018, the main contribution of our work is the data-centric analysis of recent work in this field, placing it along the mobility data continuum which ranges from detailed dense trajectories of individual movers (quasi-continuous tracking data), to sparse trajectories (such as check-in data), and aggregated trajectories (crowd information).

The GPS-driven mobile application-based ride-hailing systems, e.g., Uber and Ola, have become integral to daily life and natural transport choices for urban commuters. However, there is an imbalance between demand or pick-up requests and supply or drop-off requests in any area. The city planners and the researchers are working hard to balance this gap in demand and supply situation for taxi requests. The existing approaches have mainly focused on clustering the spatial regions to identify the hotspots, which refer to the locations with a high demand for pick-up requests. This study determined that if the hotspots focus on clustering high demand for pick-up requests, most of the hotspots pivot near the city center or in the two-three spatial regions, ignoring the other parts of the city. This paper (An earlier version of this paper was presented at the Australasian Database Conference and was published in its Proceedings: https://link.springer.com/chapter/10.1007/978-3-030-69377-0_10) presents a hotspot detection method that uses a dominating set problem-based solution in spatial-temporal space, which covers high-density taxi pick-up demand regions and covers those parts of the city with a moderate density of taxi pick-up demands during different hours of the day. The paper proposes algorithms based on k-hop dominating set; their performance is evaluated using real-world datasets and proves the edge over the existing state-of-the-art methods. It will also reduce the waiting time for customers and drivers looking for their subsequent pick-up requests. Therefore, this would maximize their profit and help improve their services.

Outlier detection and cleaning are essential steps in data preprocessing to ensure the integrity and validity of data analyses. This paper focuses on outlier points within individual trajectories, i.e., points that deviate significantly inside a single trajectory. We experiment with ten open-source libraries to comprehensively evaluate available tools, comparing their efficiency and accuracy in identifying and cleaning outliers. This experiment considers the libraries as they are offered to end users, with real-world applicability. We compare existing outlier detection libraries, introduce a method for establishing ground-truth, and aim to guide users in choosing the most appropriate tool for their specific outlier detection needs. Furthermore, we survey the state-of-the-art algorithms for outlier detection and classify them into five types: Statistic-based methods, Sliding window algorithms, Clustering-based methods, Graph-based methods, and Heuristic-based methods. Our research provides insights into these libraries’ performance and contributes to developing data preprocessing and outlier detection methodologies.