Although Vision Language Models (VLMs) exhibit strong perceptual abilities and impressive visual reasoning, they struggle with attention to detail and precise action planning in complex, dynamic environments, leading to subpar performance. Real-world tasks typically require complex interactions, advanced spatial reasoning, long-term planning, and continuous strategy refinement, usually necessitating understanding the physics rules of the target scenario. However, evaluating these capabilities in real-world scenarios is often prohibitively expensive. To bridge this gap, we introduce DeepPHY, a novel benchmark framework designed to systematically evaluate VLMs' understanding and reasoning about fundamental physical principles through a series of challenging simulated environments. DeepPHY integrates multiple physical reasoning environments of varying difficulty levels and incorporates fine-grained evaluation metrics. Our evaluation finds that even state-of-the-art VLMs struggle to translate descriptive physical knowledge into precise, predictive control.

In the existing traffic prediction scenarios, the lack of accompanying event data, noise interference and insufficient supervised signals seriously restrict the effect of actual traffic prediction. Meanwhile, currently prevalent graph neural networks often struggle to capture effective semantic structures when dealing with learning tasks involving diverse specific events, consequently exhibiting limited generalization and transfer capabilities. This study focuses on crowd gathering events in transportation scenarios and conducts quantitative analysis of their potential risks to traffic network. Relying on the massive online crowd query data produced in Location Based Services (LBS), we propose a generative strategy for node and edge augmentation based on event-traffic interactions, which seeks to generate richer supervised signals. Furthermore, in response to the generative graph structure derived from event chains that fail to match the contextual semantic information, we utilize comparative learning for self-supervised training as the auxiliary proxy task of time series prediction. Experiments on the benchmark datasets of real road networks show that the proposed method is effective in identifying the traffic risk of road segments, especially when the breakdown probability is greater than 50%.

Large Language Models (LLMs) demonstrate exceptional performance in textual understanding and tabular reasoning tasks. However, their ability to comprehend and analyze hybrid text, containing textual and tabular data, remains unexplored. The hybrid text often appears in the form of hybrid long documents (HLDs), which far exceed the token limit of LLMs. Consequently, we apply an Automated Information Extraction framework (AIE) to enable LLMs to process the HLDs and carry out experiments to analyse four important aspects of information extraction from HLDs. Given the findings: 1) The effective way to select and summarize the useful part of a HLD. 2) An easy table serialization way is enough for LLMs to understand tables. 3) The naive AIE has adaptability in many complex scenarios. 4) The useful prompt engineering to enhance LLMs on HLDs. To address the issue of dataset scarcity in HLDs and support future work, we also propose the Financial Reports Numerical Extraction (FINE) dataset. The dataset and code are publicly available in the attachments.

Cross-database knob tuning has long been recognized as a critical but complex task. Modern database systems expose hundreds of configuration knobs that play key roles in memory management, concurrency control, and query optimization. Proper tuning can significantly improve performance, while improper settings can cause severe degradation. Despite progress with black-box methods like reinforcement learning and bayesian optimization, as well as LLM-based tuning guides, challenges remain, such as modeling knob dependencies, high cold-start trial costs, and weak handling of dynamic workloads. We propose KnobTuneX, a structure-aware, LLM-enhanced framework for automatic database tuning that integrates domain knowledge, historical behaviors, and reasoning capabilities to adapt to diverse workloads. The approach features an offline learning stage to capture knob-performance relationships and build a historical RAG store, and an online inference stage that dynamically recommends knobs based on structured reasoning and historical insights. By explicitly modeling dependencies among knobs and leveraging LLMs for informed decision-making, the framework achieves both interpretability and adaptability. Finally, we evaluate KnobTuneX on PostgreSQL and show that the method outperforms mainstream approaches in efficiency, scalability, and overall tuning quality across OLTP and OLAP workloads. The implementation of our work can be found at https://github.com/vjwww/KnobTuneX.

Generating high-quality pseudo-labels on the cloud side is crucial for cloud-edge collaborative object detection, especially in dynamic traffic monitoring scenarios where the target data distribution continuously evolves. Existing methods often assume a perfectly reliable cloud model, neglecting the potential for errors in the cloud’s predictions, or employ simple adaptation techniques that struggle to handle complex distribution shifts. This paper proposes a novel Cloud-Adaptive High-Quality Pseudo-label generation algorithm (CA-HQP) that addresses these limitations by incorporating a learnable Visual Prompt Generator (VPG) and a dual feature alignment strategy into the cloud model updating process. The VPG enables parameter-efficient adaptation of the large pre-trained cloud model by injecting task-specific visual prompts into the model’s input, enhancing its flexibility without extensive fine-tuning. To mitigate domain discrepancies, CA-HQP introduces two complementary feature alignment techniques: a global Domain Query Feature Alignment (DQFA) that captures scene-level distribution shifts and a fine-grained Temporal Instance-Aware Feature Embedding Alignment (TIAFA) that addresses instance-level variations. Extensive experiments on the Bellevue traffic dataset, a challenging real-world traffic monitoring dataset, demonstrate that CA-HQP significantly improves the quality of pseudo-labels compared to existing state-of-the-art cloud-edge collaborative object detection methods. This translates to notable performance gains for the edge model, showcasing the effectiveness of CA-HQP in adapting to dynamic environments. Further ablation studies validate the contribution of each individual component (DQFA, TIAFA, VPG) and confirm the synergistic effect of combining global and instance-level feature alignment strategies. The results highlight the importance of adaptive cloud model updates and sophisticated domain adaptation techniques for achieving robust and accurate object detection in continuously evolving scenarios. The proposed CA-HQP algorithm provides a promising solution for enhancing the performance and reliability of cloud-edge collaborative object detection systems in real-world applications.

In contemporary urban management, the rapid advancement of technology and widespread use of location-based services provide critical insights, particularly in traffic planning. Analyzing urban spatiotemporal trajectory data reveals diverse perspectives that inform effective city management. This paper focuses on extracting insights from spatiotemporal trajectories and introduces a suite of urban data mining techniques applicable across scenarios, including points of interest diversification. The framework aids in understanding city dynamics, including the distribution and activities of individuals. The method begins by constructing spatiotemporal stay information through two processes: stay point and stay area extraction. Spatiotemporal semantic trajectory data is then integrated with Points of Interest (POI) semantic details. A novel clustering method based on network community detection is proposed. Our experiments show that the method not only achieves comparable or superior results but also ensures time efficiency, aligning with the demands of the big data era.

The generation of adversarial safety-critical scenarios is essential for rigorously evaluating autonomous driving systems, enabling the identification of vulnerabilities and enhancement of system robustness. However, existing methodologies predominantly focus on extreme, unconstrained collision scenarios in which non-player character (NPC) vehicles exhibit unrealistic adversarial behaviors toward the ego vehicle. While such scenarios serve as stress tests, their practical utility is limited due to two key factors: 1) these extreme events are statistically rare in real-world traffic and frequently involve collisions that are physically unavoidable, irrespective of the autonomous vehicle’s decision-making capabilities; and 2) NPC behaviors in these scenarios are often intentionally aggressive (e.g., deliberate rear-end collisions), resulting in liability attribution that predominantly lies with the NPCs rather than exposing meaningful system limitations. Recent efforts to enhance scenario plausibility rely extensively on large-scale real-world traffic datasets, introducing significant computational costs and scalability constraints. To overcome these limitations, we propose a three-layer relative safety region model that partitions the driving environment into zones of varying risk levels. This partitioning increases the likelihood that NPC vehicles will interact within relative safety boundary regions, thus enabling the generation of more realistic and contextually relevant adversarial scenarios without the need for extensive real-world traffic data. We introduce AuthSim, a platform that integrates this safety model with reinforcement learning (RL) to generate both authentic and effective safety-critical scenarios. AuthSim is the first comprehensive approach to address both the authenticity and effectiveness of autonomous driving test scenarios without relying on large-scale traffic data. Empirical results demonstrate that AuthSim outperforms existing methods, achieving a 5.25% improvement in average cut-in distance and a 11.94% increase in average collision interval time compared with the state-of-the-art (SOTA) results, all while maintaining superior efficiency in scenario generation. These findings highlight the potential of AuthSim to produce high-fidelity and efficient test cases for the rigorous evaluation of autonomous driving systems.

Sheng Yang, Yurong Wu, Yan Gao, Zineng Zhou, Bin Benjamin Zhu, Xiaodi Sun, Jian-Guang Lou, Zhiming Ding, Anbang Hu, Yuan Fang, Yunsong Li, Junyan Chen, Linjun Yang. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

The clustering method based on graph models has garnered increased attention for its widespread applicability across various knowledge domains. Its adaptability to integrate seamlessly with other relevant applications endows the graph model-based clustering analysis with the ability to robustly extract "natural associations" or "graph structures" within datasets, facilitating the modelling of relationships between data points. Despite its efficacy, the current clustering method utilizing the graph-based model overlooks the uncertainty associated with random walk access between nodes and the embedded structural information in the data. To address this gap, we present a novel Clustering method for Maximizing Decoding Information within graph-based models, named CMDI. CMDI innovatively incorporates two-dimensional structural information theory into the clustering process, consisting of two phases: graph structure extraction and graph vertex partitioning. Within CMDI, graph partitioning is reformulated as an abstract clustering problem, leveraging maximum decoding information to minimize uncertainty associated with random visits to vertices. Empirical evaluations on three real-world datasets demonstrate that CMDI outperforms classical baseline methods, exhibiting a superior decoding information ratio (DI-R). Furthermore, CMDI showcases heightened efficiency, particularly when considering prior knowledge (PK). These findings underscore the effectiveness of CMDI in enhancing decoding information quality and computational efficiency, positioning it as a valuable tool in graph-based clustering analyses.

Emergency relief operations are essential in disaster aftermaths, necessitating effective resource allocation to minimize negative impacts and maximize benefits. In prolonged crises or extensive disasters, a systematic, multi-cycle approach is key for timely and informed decision-making. Leveraging advancements in IoT and spatio-temporal data analytics, we’ve developed the Multi-Objective Shuffled Gray-Wolf Frog Leaping Model (MSGW-FLM). This multi-constraint, multi-objective resource allocation model has been rigorously tested against 28 diverse challenges, showing superior performance in comparison to established models such as NSGA-II, IBEA, and MOEA/D. MSGW-FLM’s effectiveness is particularly notable in complex, multi-cycle emergency rescue scenarios, which involve numerous constraints and objectives. This model represents a significant step forward in optimizing resource distribution in emergency response situations.

With the vigorous development of transportation infrastructure in various countries, the traffic network within the city is becoming more and more complex, and when an emergency occurs in one or more areas of the city, it will inevitably cause traffic congestion in the area and keep spreading. There are still many challenges to solve the urban emergency route planning problem. In this paper, we have employed a double layer search structure, where we have empowered the traditional A* model with a neural network, to construct a region-level dynamic path planning model known as “Double Layer A*”. The model divides the road network into two layers, and implements the outer layer and inner layer search. In the outer layer search, we use the historical cab travel data for training to achieve the general direction planning; in the inner layer search, we update the original planning according to the changes of the road condition characteristics of the regional nodes, and perform the re-planning in real time. We conducted experimental evaluations using the road network data of Beijing, and the results showed that compared to a single-layer search structure path planning model, our Double layer A* model planned paths with higher similarity in land characteristics, connectivity, and average connectivity between adjacent nodes, which demonstrates the effectiveness and reasonableness of the Double layer A* model in emergency path planning.

Traffic patterns in the spatiotemporal network are affected by temporal dynamics and spatial correlations. The network flows have different strengths interacting at various implicit layers, and this dynamic process needs to be further explored. Predicting future traffic based on historical data from transportation IoT has been well studied, however, most of the works focus on traffic dynamics in the homogeneous spatial or temporal structure. When the spatiotemporal graph structure turns complex, it becomes a challenge to capture the deep traffic patterns on it. In this paper, a heterogeneous modular flows graph is constructed to characterize the implied spatiotemporal correlations within the traffic data. Then, we proposed a Multilayer Graph Skip Temporal Convolution Network (MGSTCN) which extracts skip aggregated representations of node status to the modular flows graph. And an extended random walk on diverse modular graphs is used to learn the spatial dependencies. The experiments based on real traffic networks confirmed that the MGSTCN has a better performance compared to the spatiotemporal homogeneous methods.

DARTS search space (DSS) has become a canonical benchmark for NAS whereas some emerging works pointed out the issue of narrow accuracy range and claimed it would hurt the method ranking. We observe some recent studies already suffer from this issue that overshadows the meaning of scores. In this work, we first propose and orchestrate a suite of improvements to frame a larger and harder DSS, termed LHD, while retaining high efficiency in search. We step forward to renovate a LHD-based new benchmark, taking care of both discernibility and accessibility. Specifically, we re-implement twelve baselines and evaluate them across twelve conditions by combining two underexpolored influential factors: transductive robustness and discretization policy, to reasonably construct a benchmark upon multi-condition evaluation. Considering that the tabular benchmarks are always insufficient to adequately evaluate the methods of neural architecture search (NAS), our work can serve as a crucial basis for the future progress of NAS.

Abstract With the development of mobile networks and the rapid prevalence of location‐based social networks (LBSN), a massive volume of spatiotemporal data has been generated, which is valuable for points of interest (POI) recommendation. However, current studies have not unleashed the full power of such spatiotemporal data, which either explore only a single dimension of the data or consider multiple factors in an asynchronous fashion. In this article, we propose a novel spatiotemporal network‐based recommender framework (STNBR) to effectively recommend POIs for users. Specifically, we first establish a comprehensive conceptual model of spatiotemporal data, involving various essential factors for POIs recommendation. On top of the conceptual model, we design a series of meaningful meta‐paths that simultaneously consider the time and location factors to precisely capture the semantics of user behaviours. By profiling users based on their embedded meta‐paths, our approach can yield meaningful POIs recommendations. We have evaluated our proposal using a realistic dataset obtained from Foursquare and Gowalla, the results of which show that our STNBR model outperforms existing approaches.

The task of few-shot classification (FSC) is to build a model to discriminate novel categories that do not present in training categories with limited labeled samples. However, in many applications, the training categories and novel categories are assumed to come from the same domain. Existing few-shot classification algorithms achieve promising performance in a single domain, but often fail to generalize to unseen domains because of the different feature distribution across domains. The main goal of this work is to propose an effective recognition model that can work on various image domains with domain shift. Specifically, we propose to construct a new domain-attention mechanism with some adapters based on the squeeze-and-excitation network architectures . The proposed network processes multiple domains simultaneously and all parameters are shared across domains. In addition, because of the large discrepancy in feature distributions among different domains, similarity transformation layers are used to reduce the differences in the feature distributions in the training stage. Extensive experiments were conducted to validate the domain generalization capability (GC) of this model on four FSC datasets: mini-ImageNet, Cars, CUB-200-2011, and Simpsons Characters Data. The results show that the presented method has excellent performances on various datasets across diverse domains.

With the widespread use of the Global Positioning System (GPS) in the fields such as traffic monitoring, sports navigation, and track recording, the trajectory data recording users’ spatial and temporal information has grown dramatically. The huge volume of trajectory data causes high cost and poses a great challenge to data storage, network transmission, query and analysis. Therefore, the compression of trajectory data becomes a crucial issue. This paper proposes an online trajectory compression algorithm based on vector extraction (VOLTCom), which aims to achieve efficient data compression while retaining more effective information, and is mainly applied to trajectory recording and analysis in the traffic field. VOLTCom first generates vectors for trajectory data according to customized vector features, and then performs real-time vector extraction to achieve online trajectory compression. The vector extraction of the trajectory data ensures the stability of the compression time per unit and achieves efficient compression. Experiments on real datasets show that VOLTCom can retain the information of object velocity variation by vector density and outperforms traditional algorithms in terms of error, compression rate, and execution time. The algorithm is$O(1)$in compression time complexity and has better compression performance.

The skyline query is one of the most important supporting technologies for the location-based query services in the road network. Usually, when a user queries the skyline points in the road network, the query area is a user-centered circle or rectangle area, without considering the impact of the current movement speed and direction of the user on the formation of the query area. In this context, a speed and direction aware skyline query method is proposed, which can provide the skyline query area for the users by considering their moving speed and direction. Since the efficiency to directly obtain points of interest from speed and direction aware query area is not high, a Voronoi based speed and direction query area generation algorithm is proposed to approximate the query area, so as to improve the obtaining efficiency of points of interest in the area. The experiments on road networks and points of interest data of Beijing show the performance of the proposed method in terms of query efficiency and quality.

With the deterioration of the transportation ecosystem and traffic congestion, traffic graph representations become more complex and lack intelligibility. The deep learning method provides a new way for traffic prediction by mining the spatiotemporal relations of historical states in traffic network. Recurrent neural networks based on gate control can avoid the gradient vanishing and graph convolutional networks provide theoretical support for spatial feature extraction of traffic graph. However, in the prediction of traffic evolution behaviors, the statuses of units and groups distributed in the traffic network have strong and continuous spatiotemporal correlations. Modeling only based on the temporal or spatial perspective usually lacks comprehensiveness and semantic relevance which results in poor prediction performance. In this paper, we propose a spatiotemporal hierarchical propagation graph convolutional network(SHPGCN) to get the whole spatiotemporal evolution features of traffic graph, which can be used to predict the propagation effects and state changes of traffic flow. Considered the inadequacy of graph convolution in the learning of hierarchical features, we construct a low-dimensional propagation graph representation that projects complex node relationships into first-order neighborhoods to capture dynamic changes at different spatial and temporal scales. The SHPGCN uses graph convolution to encode the traffic propagation features and embeds them into a bidirectional recurrent sequence for traffic prediction. The experiment results show that SHPGCN can provide good prediction accuracy and robustness for the traffic evolution behaviors.

Differentiable architecture search (DARTS) is widely considered to be easy to overfit the validation set which leads to performance degradation. We first employ a series of exploratory experiments to verify that neither high-strength architecture parameters regularization nor warmup training scheme can effectively solve this problem. Based on the insights from the experiments, we conjecture that the performance of DARTS does not depend on the well-trained supernet weights and argue that the architecture parameters should be trained by the gradients which are obtained in the early stage rather than the final stage of training. This argument is then verified by exchanging the learning rate schemes of weights and parameters. Experimental results show that the simple swap of the learning rates can effectively solve the degradation and achieve competitive performance. Further empirical evidence suggests that the degradation is not a simple problem of the validation set overfitting but exhibit some links between the degradation and the operation selection bias within bilevel optimization dynamics. We demonstrate the generalization of this bias and propose to utilize this bias to achieve an operation-magnitude-based selective stop.

Differentiable architecture search (DARTS) is a promising end to end NAS method which directly optimizes the architecture parameters through general gradient descent. However, DARTS is brittle to the catastrophic failure incurred by the skip connection in the search space. Recent studies also cast doubt on the basic underlying hypotheses of DARTS which are argued to be inherently prone to the performance discrepancy between the continuous-relaxed supernet in the training phase and the discretized finalnet in the evaluation phase. We Figure out that the robustness problem and the skepticism can both be explained by the information bypass leakage during the training of the supernet. This naturally highlights the vital role of the sparsity of architecture parameters in the training phase which has not been well developed in the past. We thus propose a novel sparse-regularized approximation and an efficient mixed-sparsity training scheme to robustify DARTS by eliminating the information bypass leakage. We subsequently conduct extensive experiments on multiple search spaces to demonstrate the effectiveness of our method.

With the development of deep learning technology, a series of new recommendation algorithms are produced by using the idea of deep learning in the field of recommendation algorithm. These recommendation algorithms usually only mine user, item and context related features, and then combine the features, and use neural network for model training. However, it may be difficult for us to find a suitable feature expression for some features. At the same time, it is difficult for us to find all the feature expression forms for all kinds of features of a certain thing, which leads to the recommendation effect not reaching the expected level. To solve this problem, this paper proposes a deep neural network based on explicit and implicit feature joint (EIFJDNN) model, which is different from the features determined in Feature Engineering in the past. The features used in this model include explicit features and implicit features. In the recommendation of interest points, the access of interest points is affected by both time and space, we take on temporal and spatial features as explicit features, and other influencing factors as implicit features for recommendation. We designed FGMF algorithm, which is an improvement of GMF model. We use it to pretrain the model, then input the results into deep neural network for training, predict the score of each user on the points of interest, and then rank the scores to get the final recommendation list. Finally, we verified the recommendation performance of EIFJDNN model on Foursquare dataset, and the results show that the proposed model is superior to the existing work in hit rate and normalized discounted cumulative gain.

With the development of location-based services and smart terminals, skyline query technique has been used widely in intelligent transportation systems. In skyline queries, the areas and keywords queried by users have a great impact on the quality of the query results and users may only be interested in the closer results of the skyline query. However, current approaches for the continuous skyline query limit the area of the skyline query to a specific area in the road network, which leads to that many useful query results cannot be retrieved. To this end, an innovative continuous skyline query approach in city range is proposed in this paper, where a multi-scale area divisions of the urban road network are provided to find the optimize query scale and area. In our approach, first, the dominant area of each intersection node in the road network is established based on the Voronoi. Then, all Points of Interest ($POI\text{s}$) are divided into the dominant area of each intersection node. After that, the intersection node aggregation algorithm ($INAA$), link remolding algorithm ($LMA$) and link fitting algorithm ($LFA$) are proposed to reduce the number of intersection nodes in the road network, so as to increase the dominant area of the remaining intersection nodes and the number of POIs in these nodes. Finally, a better query scale by considering the efficiency and quality of the query is given through the studies.

Urban rail transit stations are the key nodes of urban rail transit network. Identifying and analyzing land use characteristics around urban rail transit stations can significantly contribute to urban rail transportation operation and management. Therefore, a visualization method of land use characteristics around urban rail transit stations based on POI is proposed in this paper. In the proposed method, first, the Voronoi diagram is used to determine coverage of urban rail transit stations and each POI is put in a coverage area based on their physical location. Then, topic-oriented hierarchical POIs of each urban rail transit station are extracted based on skyline idea. Finally, the land use characteristics around an urban rail transit station are visualized based on the extracted hierarchical POIs. We carried out two case studies and a quality evaluation. By using realistic data from Beijing rail transit in order to validate the method proposed in this paper. Results show that our method can clarify various situations of land use of urban rail transit stations and may provide support for the application of transportation model technology.

Accurate traffic flow prediction is becoming increasingly important for transportation planning, control, management, and information services of successful. Numerous existing models focus on short-term traffic forecasts, but effective long-term forecasting of traffic flows have become a challenging issue in recent years. To solve this problem, this paper proposes a deep learning architecture which consisting of two parts: the long short-term memory encoder-decoder structure at the bottom and the calibration layer at the top. In the encoder-decoder model, we propose an hard attention mechanism based on learning similar patterns to enhance neuronal memory and reduce the accumulation of error propagation. To correct some of the missing details, we design a control gate in the calibration layer to learn the predicted data in groups according to different forms. The proposed method is evaluated on real-world datasets and compared with other state-of-the-art methods. It is verified that our model can accurately learn local feature and long-term dependence, and has better accuracy and stability in long-term sequence prediction.

Multi-agent technologies have been widely applied to many applications, such as in e-markets, cloud computing, service-oriented environments, etc. In real applications, service-oriented environments are open and dynamic, where loosely coupled agents interact to consume and provide services. How to accurately evaluate the potential performance (i.e., QoS) of service providers on the service requested by a service consumer in such open and dynamic environments is a challenging issue in both theory and practice. In this paper, an innovative approach is proposed to evaluate the QoS of service providers in service-oriented environments. The proposed approach first borrows the reference report mechanism from the certified reputation model, so as to efficiently collect reference reports (i.e., historical performance) of service providers in open and dynamic environments. Then, a tensor-based QoS model is proposed to construct multi-dimensional relationships between QoS evaluation factors and the QoS values of service providers based on the collected reference reports. The QoS evaluation factors include the type of services, the performance of service providers, the subjectivity of service consumers, the time slot of reference reports. Finally, a CANDECOMP/PARAFAC decomposition and gradient descent-based mechanism is used to evaluate the QoS values of service providers through completing the missing entry values in the constructed tensor. The uniform random simulation experiments indicate that the proposed approach can achieve efficient and accurate QoS evaluation in service-oriented environments with only limited collected reference reports, especially when some service providers do not have reference reports.

Processing application user reviews has recently been recognized as an efficient approach to explore user requirements. However, most existing approaches focus on mining the reviews themselves without effectively associating the reviews with requirements concepts, limiting the effectiveness of review mining for requirements analysis tasks. In this paper, we propose to automatically identify Requirements-oriented Reviews (RoRs) from software application reviews by considering requirements specific domain knowledge and syntactic information of user reviews. Specifically, we first define a conceptual model of RoRs based on existing requirements ontology and user review categories, establishing connections between the concepts of requirements engineering and user reviews. We then systematically identify the textual features of RoRs by following a conceptual model-driven top-down strategy. Based on such features, we then train effective RoR classifiers to identify RoRs. To evaluate the performance of our approach, we have applied our approach to a real dataset of mobile application reviews, the results of which show that our approach can effectively identify RoRs with an F-measure of 0.8, outperforming than the baselines.

Summary Regional population density has temporal and spatial characteristics, and most of the existing prediction models fail to take these two characteristics into account at the same time, which results in unsatisfactory forecasting results. To address this problem, we use the deep learning models to predict the crowd distribution in the evacuation area, so as to realize the recommendation of the evacuation area. First, a raster population density prediction model based on long short‐term memory (LSTM) is studied, and then a multiarea population density prediction model considering temporal and spatial characteristics, named ST‐LSTM, is designed. The results of our extensive experiments on the real dataset show that our proposed ST‐LSTM is both effective and efficient.

In recent years, the amount of user check-in data has significantly increased on social network platforms. Such data is an ideal source for characterizing user behaviors and identifying similar users, contributing to many research areas (e.g. user-based collaborative filtering). However, existing trajectory-based user similarity analysis approaches do not distinguish the effects of geographical factors at a fine-grained level, and thus are not able to unleash the full power of semantic information that is hidden in the trajectory. In this paper, we have proposed an effective graph embedding approach to identify similar users based on their check-in data. Specifically, we firstly identify meaningful concepts of user check-in data, based on which we design two metagraphs for representing features of similar user behaviors. Then we characterize each user with a sequence of nodes that are derived through a metagraph-guided random walk strategy. Such sequences are embedded to generate meaningful user vectors for measuring user similarity and eventually identifying similar users. We have evaluated our proposal on three public datasets, the results of which show that our approach is 4% higher than the best existing approach in terms of F1-measure.

With the development of society and economy, the urban rail transit has become one of the important components of urban transportation system, while the construction of the urban rail greatly improves the public transportation environments. Currently, there are many research focus on the passenger flow predictions according to their corresponding historical data, however, it is hard to assist transport models vary such volumes for a new station planning or being constructed. In view of this limitation, we provide a novel method for urban rail station characteristics analysis in intelligent transportation considering city land usages. Initially, point of interest (POIs) are divided by the proposed RC-tree (Colored R-tree)-based algorithm into the bounded areas for each station. Second, the Diversity and Proportion approaches are proposed to extract the top-k POIs from bounded areas based on their semantic and spatial characteristics. Then, classify the stations based on the similarity of the extracted top-k POIs. Moreover, we made a case study on real dataset, including a large volume of Automatic Fare Collection system (AFC) records for the experimental evaluations, and the results show that the proposed method can verify the rationality of land use and provide support for the application of transportation model technology.

Eliciting precise user preferences and establishing a comprehensive user profile significantly contribute to personalized recommendations from numerous applications. However, existing methods do not adequately establish critical relationships at the knowledge level. In this paper, we argue that precisely processing applications' user feedback is essential for understanding user requirements and making application recommendations. Therefore, we first establish an ontological model of user feedback, guiding the generation of a knowledge graph regarding user reviews and user ratings. In particular, we augment the graph with topics of each review and application in order to deal with the sparsity of user feedback. Moreover, we explore in-depth knowledge from the graph by identifying three meaningful meta-paths, which are essential for calculating user similarity and thus making recommendations. Specifically, we propose a feedback-based similarity calculation model FSCM, with the purpose of predicting applications that are of interests of certain users. We have evaluated our model over 1386 reviews from a mobile application store, the results of which show that our approach can improve the prediction accuracy, as well as to enhance the interpretability of analysis results.

With the developing of smart sensors and mobile devices produces an increasing volume of data, and it captures the states of transportation infrastructures. Such data are collected and uploaded frequently, which forms the heavy data calculation and storage. Moreover, state of monitored object may be keeping the same or slight change according to a certain state during a period, such as moving vehicles on a certain path with a basic uniform speed. Therefore, if trajectory pattern of vehicles can be obtained through the state change mode, scale and update frequency of the data can be greatly reduced. Based on the above-mentioned ideas, we are aware that the trajectory data storage is divided into traceability storage and vector storage, where original sampled data from sensing device, and state vectors are extracted from the analysis of the original sample data. In this way, only a relatively small amount of vector data is stored. The system will not only effectively reduce the frequency of sampling data storage, but also reduce query and analysis operations involved with the amount of data. The vector function is used to represent road network with indexes, and the data query based on the road network is used to extract the semantic information. Our experimental results show that our proposed methodologies have significant improvements in intelligent transportation.

We are witnessing increasing interests in developing “smart cities” which helps improve the efficiency, reliability, and security of a traditional city. An important aspect of developing smart cities is to enable “smart transportation,” which improves the efficiency, safety, and environmental sustainability of city transportation means. Meanwhile, the increasing use of GPS devices has led to the emergence of big trajectory data that consists of large amounts of historical trajectories and real-time GPS data streams that reflect how the transportation networks are used or being used by moving objects, e.g., vehicles, cyclists, and pedestrians. Such big trajectory data provides a solid data foundation for developing various smart transportation applications, such as congestion avoidance, reducing greenhouse gas emissions, and effective traffic accident response, etc. Instead of proposing yet another specific smart transportation application, we propose the parallel-distributed network-constrained moving objects database (PD-NMOD), a general framework that manages big trajectory data in a scalable manner, which provides an infrastructure that is able to support a wide variety of smart transportation applications and thus benefiting the smart city vision as a whole. The PD-NMOD manages both transportation networks and trajectories in a distributed manner. In addition, the PD-NMOD is designed to support general SQL queries over moving objects and to efficiently process the SQL queries on big trajectory data in parallel. Such design facilitates smart transportation applications to retrieve relevant trajectory data and to conduct statistical analyses. Empirical studies on a large trajectory data set collected from 3,500 taxis in Beijing offer insight into the design properties of the PD-NMOD and offer evidence that the PD-NMOD is efficient and scalable.

When planning routes, drivers usually consider a multitude of different travel costs, e.g., distances, travel times, and fuel consumption. Different drivers may choose different routes between the same source and destination because they may have different driving preferences (e.g., time-efficient driving v.s. fuel-efficient driving). However, existing routing services support little in modeling multiple travel costs and personalization-they usually deliver the same routes that minimize a single travel cost (e.g., the shortest routes or the fastest routes) to all drivers. We study the problem of how to recommend personalized routes to individual drivers using big trajectory data. First, we provide techniques capable of modeling and updating different drivers' driving preferences from the drivers' trajectories while considering multiple travel costs. To recommend personalized routes, we provide techniques that enable efficient selection of a subset of trajectories from all trajectories according to a driver's preference and the source, destination, and departure time specified by the driver. Next, we provide techniques that enable the construction of a small graph with appropriate edge weights reflecting how the driver would like to use the edges based on the selected trajectories. Finally, we recommend the shortest route in the small graph as the personalized route to the driver. Empirical studies with a large, real trajectory data set from 52,211 taxis in Beijing offer insight into the design properties of the proposed techniques and suggest that they are efficient and effective.

Tracking and managing the locations of moving objects are essential in modern intelligent transportation systems (ITSs). However, a number of limitations in existing methods make them unsuitable for real-world ITS applications. In particular, Euclidean-based methods are not accurate enough in representing locations and in analyzing traffic, unless the locations are frequently updated. Network-based methods require either digital maps to be installed in moving objects or transmission of prediction policies, which inevitably increase the cost. To solve these problems, we propose a network-matched trajectory-based moving-object database (NMTMOD) mechanism and a traffic flow analysis method using the NMTMOD. In the NMTMOD, the locations of moving objects are tracked through a dense sampling and batch uploading strategy, and a novel edge-centric network-matching method, which is running at the server side, is adopted to efficiently match the densely sampled GPS points to the network. In addition, a deviation-based trajectory optimization method is provided to minimize the trajectory size. Empirical studies with large real trajectory data set offer insight into the design properties of the proposed NMTMOD and suggest that the NMTMOD significantly outperforms other mobile-map free-moving-object database models in terms of precision of both location tracking and network-based traffic flow analysis.

In the big data era, the principal traffic bottleneck in a data center is the communication between hosts. Moreover, the redundant packet contents indeed take a major part of the bottleneck. In this paper, we propose a distributed and efficient cache sharing protocol, named DCP, to eliminate redundant traffic in data center with Fat-Tree topology. In DCP, each switch holds packet caches, which are used for eliminating redundant packets. A cache sharing mechanism using Bloom Filter is also proposed to share local caches of switches with servers. Moreover, we propose another technique leveraging Bloom Filter to reduce false positive ratio caused by sharing cache information between switches and servers. We build up a Fat-Tree topology with k = 16 to evaluate the performance of DCP, and the simulation results show that DCP can eliminate 40% to 60% redundant packets, which validate its efficiency.

Map Generalization is one of the most fundamental technologies for modern digital maps. It can effectively reduce the storage space and fit to different applications according to their scale requirement. This paper presents an efficient solution for this problem that won the ACM SIGSPATTAL CUP 2014. Given the original geometries which are represented by sampling points sequence, this method divides the boundaries into many small segments based on their topological characteristics and constriants. It attempts to minimize the number of sampling points by simplifying the given map and constraining points. In addition, the method also employs many optimization techniques to reduce the total latency, like memory pool, parallel computing and string parsing. Experimental results on real datasets demonstrate the effectiveness and efficiency of the proposed method.

The availability of massive trajectory data collected from GPS devices has received significant attentions in recent years. A hot topic is trip recommendation, which focuses on searching trajectories that connect (or are close to) a set of query locations, e.g., several sightseeing places specified by a traveller, from a collection of historic trajectories made by other travellers. However, if we know little about the sample coverage of trajectory data when developing an application of trip recommendation, it is difficult for us to answer many practical questions, such as 1) how many (future) queries can be supported with a given set of raw trajectories? 2) how many trajectories are required to achieve a good-enough result? 3) how frequent the update operations need to be performed on trajectory data to keep it long-term effective? In this paper, we focus on studying the overall quality of trajectory data from both spatial and temporal domains and evaluate proposed methods with a real big trajectory dataset. Our results should be useful for both the development of trip recommendation systems and the improvement of trajectory-searching algorithms.

Mobile phones with a rich set of embedded sensors have been applied in various collaborative sensing applications. In some applications, to encourage mobile phone users performing collaborative sensing tasks, the data demanders may pay mobile phone users. However, none of the existing works takes into account it. In this paper, we study the Minimum Cost of Attaining the Required Data with mobile phones (MCARD) problem in collaborative sensing network. Given sensing regions R = {R1, R2, ..., Rm}, the set of requisite data Difor each sensing region Riand a set of mobile phones M, the MCARD problem is how to select mobile phones to get all the required data such that the total cost on paying mobile phone users is minimized. We first formally define the MCARD problem. Then, we propose an approximation algorithm for the MCARD problem with the determinate trajectories of mobile phones and a heuristic algorithm for that trajectories are unknown respectively. Simulation results demonstrate our algorithms are efficient.

This paper introduces two new coverage problems in mission-driven camera sensor networks, namely the target temporal effective-sensing coverage with non-adjustable cameras (TEC-NC) problem and the target-temporal effective-sensing coverage with adjustable cameras (TEC-AC) problem. Given a mission period, the objective of the problems is to find a sleep-wakeup schedule of the camera sensor nodes such that the overall target-temporal coverage is maximized. We formally introduce a method called Identifiability Test to check if a target with a face direction is effectively-covered by a camera sensor, and prove the problems are NP-hard. For TEC-NC, we propose a 2-approximation algorithm and two heuristic algorithms. We also design a greedy strategy which can be combined with our solutions for TEC-NC to solve TEC-AC. The simulation results indicate the quality of the outputs of our algorithms are much better than that of the existing alternative as well as close to the theoretical optimum on average.

Trajectory search plays an important role in various applications such as trip planning and recommendation. However, most existing studies only focus on spatial proximity but ignore individual users' preferences. For example, it is inappropriate to recommend a route containing gravel roads to travelers without off-road vehicles. To accommodate various user preferences, we develop MOIR/UOTS, a trip recommendation system that supports User Oriented Trajectory Search (UOTS) [5], [7] based on our previous MOIR platform [2], [4]. Meanwhile, MOIR/UOTS also functions as a test-bed system for exploring and evaluating novel trajectory searching algorithms. In particular, we demonstrate how ordinary users can interact with MOIR/UOTS to search for trajectories with their preferences, and how MOIR/UOTS helps researchers to evaluate the performance of various algorithms.

SUMMARY Resource management has been recognised as an important topic for business process execution for a long time. Most existing works on resource allocation for business processes simply assume that the structure of a business process is always fixed, and therefore do not discuss the possibility of optimising resource allocation by adapting process structures to actual resource situations. To fill this gap, we propose a resource optimisation approach of improving process structures according to resource situations and thereby pursuing the best resource utilisation efficiency. This approach comprises a role‐based business process model for resource allocation and the strategies for optimising resource allocation in conjunction with a business process improvement. A set of heuristic rules are established to guide the resource allocation for the purposes of preventing resource conflicts, shortening the total execution time, minimising the total cost, etc. Particular algorithms are also developed to implement the resource allocation according to these rules. In addition, an experimental study is conducted to discuss the incorporation of business process improvement into resource allocation for optimal process execution. Copyright © 2012 John Wiley & Sons, Ltd.

Efficient retrieval of dynamically generated, spatial-temporal, and heterogeneous sampling data in the Internet of Things is a key challenge in recent years. However, current search engine techniques are not suitable for such kind of data. To solve this problem, we propose a Hybrid Search Engine Technique for the Internet of Things based on Spatial-Temporal, Value-based, and Keyword-based Conditions ("IoT-SVK Search Engine"), in this paper. The IoT-SVK search engine has satisfactory performances in supporting multi-modal retrieval conditions, and thus provides a good solution for real-time searching of massive sampling data in the Internet of Things.

The effectiveness of map-matching algorithms highly depends on the accuracy and correctness of underlying road networks. In practice, the storage capacity of certain hardware, e.g. mobile devices and embedded systems, is sometimes insufficient to maintain a large digital map for map-matching. Unfortunately, most existing map-matching approaches consider little about this problem. They only apply to environments with information-rich maps, but turn out to be unacceptable for map-matching on simplified road networks. In this paper, we propose a novel map-matching algorithm called Passby to work on most simplified road networks. The storage size of a digital map in disk or memory can be greatly reduced after the simplification. Even under the most simplified situation, i.e., each road segment only consists of a couple of intersection points and omits any other information of it, the experimental results on real dataset show that our Passby algorithm significantly maintains high matching accuracy. Benefiting from the small size of map, simple index structure and heuristic foresight strategy, Passby improves matching accuracy as well as efficiency.

No abstract available.

Location update strategy is one of the most important factors that affect the performance of moving objects databases. However, current motion vector based location tracking methods are designed for regular movements and are thus not suitable for transportation networks with changeful traffic conditions. To solve this problem, we propose a new location update mechanism, Adaptive Network-constrained moving object Location Update Mechanism (ANLUM), in this paper. In ANLUM, the moving object can switch between different location tracking policies according to difference traffic conditions, so that the overall performance can be improved. To evaluate the performance of the proposed method, an experimental system is implemented and the results show that ANLUM can effectively reduce the communication costs with location tracking accuracy guaranteed in traffic jammed transportation networks.

Floating Car Data (FCD) provides an economic complement to infrastructure-based traffic monitoring systems. Based on our previous MOIR platform [5], we use FCD as the data source for large-scale real-time traffic monitoring. This new function brings a challenge of efficiently handling of streaming data from a very large number of moving objects. Server overload problems can occur when a system fails to process data and queries in real-tme, which can lead to critical issues such as unbounded delay accumulation, lost monitoring accuracy or lack of spontaneity. These problems can be addressed by adopting suitable load dropping decisions. In this work, we demonstrate several load shedding techniques, focusing on decision-making based on data attributes. With the end results being quantified and visualized using real data for a large city, this proof-of-concept system provides a convincing way of validating our ideas.

The uncertainty management problem for moving objects databases has been well studied recently, with many models and algorithms proposed. However, very limited work has dealt with the index of uncertain trajectories for a running moving objects database. In this paper, we propose an index framework, the UTR- Tree, for indexing the full uncertain trajectories of network constrained moving objects. Through a dynamic index maintenance technique which is associated with location updates, the UTR-Tree can deal with the full uncertain trajectories, which include not only the historical locations of moving objects, but also their current and near future location information with uncertainty considered, so that the queries on the whole life span of the moving objects can be efficiently supported. The experimental results show that the UTR-Tree outperforms previously proposed network- based moving object index methods in dealing with full uncertain trajectories.

MOIR is a Web-based prototype to support a number of novel applications with network-constrained moving object management. Technical aspects of MOIR range from data acquisition and trajectory smoothing, trajectory data management and query processing, movement predications, and location based Web page recommendation. The demo is in the context of detailed digital road maps with 38,0000 road segments and 55,000 road intersection points, and real spatiotemporal data of over ten thousand taxis in Beijing.

Over the past decade, community structure, a statistical property of networked systems such as social network and World Wide Web, has attracted considerable attention in data mining field because it enables description and prediction of complex networks. Many highly sensitive graph clustering algorithms were developed for identification of communities having dense connections internally and loose connections with others. In this context, Newman and Girvan proposed modularity Q score for quantifying the strength of community structure and measuring the fitness of a division. The Q function has become an important standard recently. In this paper, combining the strengths of the Q score and multilevel paradigm first developed for graph partitioning, we introduced a scalable algorithm MOME (i.e. modularity-based multilevel graph clustering) to efficiently discover communities from a network. The experimental results indicated that MOME ran extremely faster and finally achieved a division with a slightly higher Q score against the latest modularity-based method and its variants, particularly when the network was of a large-scale.

The goal of SECONDO is to provide a "generic" database system frame that can be filled with implementations of various DBMS data models. SECONDO was intended originally as a platform for implementing and experimenting with new kinds of data models, especially to support spatial, spatio-temporal, and graph database models. We now feel, SECONDO has a clean architecture, and it strike a reasonable balance between simplicity and sophistication. Since all the source code is accessible and to a large extent comprehensible for students, we believe it is also an excellent tool for teaching database architecture and implementation concepts. SECONDO runs on Windows, Linux, and Solaris platforms, and consists of three major components SECONDO kernel, optimizer, and graphical user interface.

Replication is one of the key technologies in promoting the performance of mobile database systems. In this paper, a novel mobile database replication scheme, the transaction-level result-set propagation (TLRSP) model, is put forward. A conflict detection and resolution strategy based on TLRSP is discussed in detail and its implementation algorithm is proposed. In the TLRSP model, mobile users are allowed to access local replicas of the database and to submit local transactions when the system is disconnected. The locally committed transactions are sent to a fixed database server for conflict reconciliation and result-set incorporation when the system is reconnected. The TLRSP model uses the incremental refreshing method to synchronize database replicas and to maintain the consistency of the replicated mobile database system.