Large Language Models (LLMs) have demonstrated impressive ability in generation and reasoning tasks but struggle with handling up-to-date knowledge, leading to inaccuracies or hallucinations. Retrieval-Augmented Generation (RAG) mitigates this by retrieving and incorporating external knowledge into input prompts. In particular, due to LLMs' context window limitations and long-context hallucinations, only the most relevant "chunks" are retrieved. However, current RAG systems face three key challenges: (1) chunks are often retrieved independently without considering their relationships, such as redundancy and ordering; (2) the utility of chunks is non-monotonic, as adding more chunks can degrade quality; and (3) retrieval strategies fail to adapt to the unique characteristics of different queries. To overcome these challenges, we design a cost-constrained retrieval optimization framework for RAG. We adopt a Monte Carlo Tree Search (MCTS) based strategy to find the optimal chunk combination order, which considers the chunks' correlations. In addition, to address the non-monotonicity of chunk utility, instead of treating budget exhaustion as the termination condition, we design a utility computation strategy to identify the optimal chunk combination without necessarily exhausting the budget. Furthermore, we propose a configuration agent that predicts optimal configurations for each query domain, improving our framework's adaptability and efficiency. Experimental results demonstrate up to a 30% improvement over baseline models, highlighting the framework's effectiveness, scalability, and suitability. Our source code has been released at https://github.com/wang0702/CARROT.

With the rise of AI in education, knowledge tracing (KT) has become important for modeling students’ knowledge from interaction data. However, existing methods still face three major challenges, including limited modeling of personalized exercise–concept relations, low robustness to noisy interactions, and inefficient training due to suboptimal data selection. To address these issues, we propose R 2 GCurL, a novel KT framework with two key designs. First, we recast KT as a graph classification problem and construct dynamic graphs from student responses, enabling the model to capture structural relations between exercises and concepts for more personalized KT. Second, we introduce a data-centric curriculum learning strategy based on dynamic graph entropy. Under our definition, pairwise dynamic graph entropy measures graph-transition continuity, where larger values indicate stronger structural similarity. Its sequence-level aggregation is used to derive a structure-aware difficulty signal for sample scheduling. On top of this, an RL-based scheduler further adapts batch selection based on model feedback and is especially beneficial under noisier and more unstable training regimes. Theoretical analysis shows that R 2 GCurL has lower computational complexity than existing graph-based KT models. Extensive experiments on five real-world datasets confirm its effectiveness, robustness, and generalizability, including as a plug-and-play enhancement for sequence-based KT models.

Next Point-of-Interest (POI) prediction is important to various human mobility applications, such as route planning and location-based advertising. To address the spatial-temporal sparsity issues arising from users' irregular and inconsistent visit times to different POIs, multi-granular structures can be incorporated to enhance feature representation through hierarchical relationships. However, existing methods often fall short in capturing the comprehensive multi-granularity spatial-temporal correlations due to three primary limitations: (1) users' complex mobility patterns entangled in single trajectory data, (2) limited mobility patterns details due to independent modeling at each granularity, and (3) low inference efficiency in cascaded multi-granularity predictions. To tackle these challenges, we propose a novel approach that models transformations across different granularities in both spatial regions and temporal periods as a diffusion process, leading to the development of the Multi-Granularity Spatial-Temporal Diffusion Network (MGSTDN). In particular, this model adopts a multi-task architecture, where predictions at varying spatial-temporal granularities (i.e., different diffusion steps) are treated as distinct tasks. By employing a multi-granularity diffusion mechanism in both spatial and temporal dimensions, it captures more nuanced spatial-temporal correlations, enhancing the physical constraints and behavioral pattern dependencies across granularities. During the diffusion process's forward stage, coarser-grained regions and periods are derived based on fine-grained features. In the reverse stage, finer-grained regions and periods are recovered from coarse-grained features, guided by encoded historical trajectory information, until the next POI is determined. To improve computational efficiency, we introduce a multi-granularity mapping propagation matrix, enabling parallel computation and accelerating the prediction process across different granularities. We evaluated the effectiveness of MGSTDN through extensive experiments on three datasets, demonstrating significant improvements over existing methods.

Trajectory similarity is a cornerstone of trajectory data management and analysis. Traditional similarity functions often suffer from high computational complexity and a reliance on specific distance metrics, prompting a shift towards deep representation learning in Euclidean space. However, existing Euclidean-based trajectory embeddings often face challenges due to the triangle inequality constraints that do not universally hold for trajectory data. To address this issue, this paper introduces a novel approach by incorporating non-Euclidean geometry, specifically hyperbolic space, into trajectory representation learning. We present the first-ever integration of hyperbolic space to resolve the inherent limitations of the triangle inequality in Euclidean embeddings. In particular, we achieve it by designing a Lorentz distance measure, which is proven to overcome triangle inequality constraints. Additionally, we design a model-agnostic framework LH-plugin to seamlessly integrate hyperbolic embeddings into existing representation learning pipelines. This includes a novel projection method optimized with the Cosh function to prevent the diminishment of distances, supported by a theoretical foundation. Furthermore, we propose a dynamic fusion distance that intelligently adapts to variations in triangle inequality constraints across different trajectory pairs, blending Lorentzian and Euclidean distances for more robust similarity calculations. Comprehensive experimental evaluations demonstrate that our approach effectively enhances the accuracy of trajectory similarity measures in state-of-the-art models across multiple real-world datasets. The LH-plugin not only addresses the triangle inequality issues but also significantly refines the precision of trajectory similarity computations, marking a substantial advancement in the field of trajectory representation learning.

Trajectory similarity measure is a fundamental component in trajectory databases, supporting many down-stream trajectory tasks. Existing similarity functions often exhibit unacceptable time complexities, hampering their efficiency for real-world scenarios. To address this limitation, learning-based approximation techniques utilizing trajectory embeddings have been proposed. However, creating a robust embedding model presents challenges, including the lack of direct involvement in the computational similarity process, adherence to non-metric similarity spaces, and the integration of precise similarity computation alignments. To address these challenges, we introduce DTisT, a novel embedding framework that enhances trajectory embeddings by pairwise learning from dual-trajectory input models. DTisT not only captures the dynamics of trajectory similarity computation through a dual-trajectory learning model but also integrates a learnable virtual trajectory to align the embedding space with non-metric similarity spaces effectively. Additionally, we incorporate aligned information from actual similarity computations into our embedding process using an attention mask mechanism. To ensure effective learning, we adopt a pre-train and fine-tune strategy, utilizing contrastive learning during the pre-training stage. Extensive experiments conducted on two real datasets demonstrate that DTisT surpasses state-of-the-art methods, showcasing its effectiveness in trajectory similarity embedding.

Data imputation (DI) is a common means of enhancing data quality. To adapt to the flourishing field of machine learning (ML), an innovative class of imputation methods that consider downstream models in the imputation process has been proposed, denoted as DI for ML. A critical challenge within this context is establishing the optimal order for imputing a set of incomplete samples. To address this, we propose an iterative approach that strategically determines the imputation order based on the potential impact on model performance. At first, we design the impact score in a what-if manner to evaluate the significance of each incomplete data point for downstream ML models. In addition, to tackle the challenge of insufficient reliable complete data in real-world scenarios, we ingeniously leverage meta-learning mechanisms to enhance the robustness of the impact score computation. Finally, to avoid the risk of converging to local optima and non-diverse data selection during iterative imputation, we introduce a real-time feedback strategy using the Multi-Armed Bandit mechanism. By balancing immediate rewards with long-term strategic gains, our approach effectively navigates the complex optimization landscape, leading to globally optimal imputation orders. We experimentally validated our method on eight real-world datasets and five types of ML models, with the results indicating that the imputation order optimized by our method outperforms the current state-of-the-art methods.

The drug recommendation task aims to provide effective and safe prescription decision support for clinical treatment based on patients' past Electronic Health Records (EHR). However, the prevalent phenomenon of missing views in multi-source heterogeneous EHR data may cause performance degradation. This is due to the lack of sufficient information and increased learning difficulties, which limit the practical effectiveness of drug recommendation models in medical applications. In this paper, we emphasize the problems of incompleteness in practical drug recommendation and propose the Incomplete Multi-View Drug Recommendation model via Multi-Level Representation Learning and Curriculum Learning named IMDR. In particular, IMDR employs a Multi-Level Representation Learning architecture equipped with a Medical Code-Level Drug Knowledge Infusion Module and a Visit-Level Cross-View Information Module for patient representation learning to overcome the information loss caused by incomplete data. And then, a Gaussian-guided curriculum learning strategy is proposed to assist the learning process of IMDR with a novel difficulty measure to achieve effective progressive learning under missing medical views. Systematic evaluation on two large-scale real-world medical datasets, MIMIC-III and MIMIC-IV, demonstrates that IMDR reduces the Drug-Drug Interaction (DDI) rate by 2.97% compared to existing state-of-the-art drug recommendation baselines, while achieving significant improvements of 3.29% and 1.97% in Jaccard similarity scores and F1 score, respectively. Furthermore, compared to advanced incomplete multi-view learning (IML) models, IMDR's advantages in Jaccard similarity scores and F1 score further expand to 4.03% and 2.41%.

In the realm of recommender systems, accurately predicting Click-Through Rate (CTR) is a critical task that involves learning user-item interaction features. Many researchers propose novel models to mine interaction signals, but they neglect that redundancy itself causes high computational cost and leads to suboptimal performance. Some tried to remove redundancy by dropping useless features, or shrinking the size of embedding table. However, current feature selection methods are vulnerable to training stochasticity and data dynamics, while embedding size assignment techniques neglect the importance relationships between feature fields. The simple combination of the two optimization ways will also yield poor performance due to the inherent gap in their optimization targets. Hence, there is no effective paradigm that can optimize feature fields from the two aspects in a simultaneous and coordinated way. In this paper, we identify the core issue as the lack of a practical score to measure the contribution of feature fields, and propose a distribution-based field optimization framework that adopts importance distribution to provide a comprehensive view for both methods. We innovatively design a learner for each field to acquire the stable and comprehensive importance situation. Then, based on this, we eliminate noise features, and assign adaptive embedding sizes for different feature fields according to the similarity of importance. With this field optimization, our proposed framework has extremely low pre-training overhead, greatly reduces training and inference time, and even achieves more accurate prediction results with fewer feature fields.

Last-mile Delivery Parcel Volume (LDPV) quantifies the number of parcels destined for a specific region, particularly a manually divided Area-Of-Interest (AOI). Accurate prediction of AOI-level LDPV is crucial for the efficient management of logistics resources. However, the straightforward adaptation of existing prediction models often falls short, primarily due to (I) a lack of consideration for the intuition behind AOI divisions, and (II) a reliance solely on fully observed historical data, which may not inform future trends. To overcome the above pitfalls, leveraging rich AOI data and advanced parcel travel time estimation services in JD Logistics, this paper introduces a novel framework called Dual-view Prediction Networks (DualPNs). It combines a Vector-Quantified AutoEncoder (VQ-AE) and a Template-Augmented Zero-Inflated Poisson (TA-ZIP), enabling both point and probabilistic distribution predictions of AOI-level LDPV. Specifically, VQ-AE utilizes a vector quantization technique to distill a large number of AOIs into representative templates, thereby addressing the first pitfall. Subsequently, TA-ZIP dynamically integrates fully observed and lookahead features, aligning them with template-specific decoders to parameterize the probabilistic distributions, thus resolving the second pitfall. We conduct extensive experiments in two cities, comprising over 47,000 and 126,000 AOIs respectively, to demonstrate the superiority of our DualPNs over other baselines. Moreover, a real-world case study highlights the effectiveness of DualPNs for enhancing downstream courier allocation by yielding an average improvement of 1.51% in the on-time delivery rate.

Online map matching is a fundamental problem in location-based services, aiming to incrementally match trajectory data step-by-step onto a road network. However, existing methods fail to meet the needs for efficiency, robustness, and accuracy required by large-scale online applications, making this task still challenging. This paper introduces a novel framework that achieves high accuracy and efficient matching while ensuring robustness in handling diverse scenarios. To improve efficiency, we begin by modeling the online map matching problem as an Online Markov Decision Process (OMDP) based on its inherent characteristics. This approach helps efficiently merge historical and real-time data, reducing unnecessary calculations. Next, to enhance robustness, we design a reinforcement learning method, enabling robust handling of real-time data from dynamically changing environments. In particular, we propose a novel model learning process and a comprehensive reward function, allowing the model to make reasonable current matches from a future-oriented perspective, and to continuously update and optimize during the decision-making process based on feedback. Lastly, to address the heterogeneity between trajectories and roads, we design distinct graph structures, facilitating efficient representation learning through graph and recurrent neural networks. To further align trajectory and road data, we introduce contrastive learning to decrease their distance in the latent space, thereby promoting effective integration of the two. Extensive evaluations on three real-world datasets confirm that our method significantly outperforms existing state-of-the-art solutions in terms of accuracy, efficiency and robustness.

The proliferation of 3D scanning technology, particularly within autonomous driving, has led to an exponential increase in the volume of Point Cloud (PC) data. Given the rich semantic information contained in PC data, deep learning models are commonly employed for tasks such as object queries. However, current query systems that support PC data types do not process queries on semantic information. Consequently, there is a notable gap in research regarding the efficiency of invoking deep models for each PC data query, especially when dealing with large-scale models and datasets. To address this issue, this work aims to design an efficient approximate approach for supporting PC analysis queries, including PC retrieval and aggregate queries. In particular, we propose a novel framework that delivers approximate query results efficiently by sampling core PC frames within a constrained budget, thereby minimizing the reliance on deep learning models. This framework is underpinned by rigorous theoretical analysis, providing error-bound guarantees for the approximate results if the sampling policy is preferred. To achieve this, we incorporate a multi-agent reinforcement learning-based approach to optimize the sampling procedure, along with an innovative reward design leveraging spatio-temporal PC analysis. Furthermore, we exploit the spatio-temporal characteristics inherent in PC data to construct an index that accelerates the query process. Extensive experimental evaluations demonstrate that our proposed method, MAST, not only achieves accurate approximate query results but also maintains low query latency, ensuring high efficiency.

Geospatial Knowledge Graphs (KGs) are widely used data structures that integrate rich knowledge from multi-source databases and play a crucial role in applications such as data retrieval and urban management. However, existing methods for Geospatial Knowledge Graph Completion (KGC) heavily rely on extensive labeled data and lack the ability to direct inference on new unlabeled geospatial databases and thus limiting their practical deployment. To address this limitation, this paper first formalizes a novel zero-shot transfer scenario and then proposes an innovative geospatial multimodal large language model framework capable of efficient Geospatial KGC with robust zero-shot generalization capabilities. Specifically, to enable effective direct inference under significant data discrepancies inherent in zero-shot scenarios, we introduce large language models (LLMs) into the geospatial KGC problem for the first time and redefine the multimodal data processing paradigm for geospatial LLMs. Next, to overcome the challenge that LLMs cannot directly handle geospatial data, we innovatively propose a Pretrain Geospatial Encoder that performs self-supervised pretraining exclusively on spatial data. Additionally, to integrate geospatial and textual modalities, we design an adaptation component that injects geospatial features into the LLMs and introduce a multi-task fine-tuning procedure. Lastly, to ensure robustness across multi-target domain scenarios, we present an implicit data alignment strategy based on adversarial learning. Extensive evaluations conducted on four real-world datasets demonstrate that our method significantly outperforms state-of-the-art approaches in terms of accuracy and robustness.

En Route Travel Time Estimation (ER-TTE) aims to learn driving patterns from traveled routes to achieve rapid and accurate real-time predictions. However, existing methods ignore the complexity and dynamism of real-world traffic systems, resulting in significant gaps in efficiency and accuracy in real-time scenarios. Addressing this issue is a critical yet challenging task. This paper proposes a novel framework that redefines the implementation path of ER-TTE to achieve highly efficient and effective predictions. Firstly, we introduce a novel pipeline consisting of a Decision Maker and a Predictor to rectify the inefficient prediction strategies of current methods. The Decision Maker performs efficient real-time decisions to determine whether the high-complexity prediction model in the Predictor needs to be invoked, and the Predictor recalculates the travel time or infers from historical prediction results based on these decisions. Next, to tackle the dynamic and uncertain real-time scenarios, we model the online decision-making problem as a Markov decision process and design an intelligent agent based on reinforcement learning for autonomous decision-making. Moreover, to fully exploit the spatio-temporal correlation between online data and offline data, we meticulously design feature representation and encoding techniques based on the attention mechanism. Finally, to improve the flawed training and evaluation strategies of existing methods, we propose an end-to-end training and evaluation approach, incorporating curriculum learning strategies to manage spatio-temporal data for more advanced training algorithms. Extensive evaluations on three real-world datasets confirm that our method significantly outperforms state-of-the-art solutions in both accuracy and efficiency.

Traffic accidents pose a significant risk to human health and property safety. To address this issue, predicting their risks has garnered growing interest. We argue that a desired prediction solution should demonstrate resilience to the complexity of traffic accidents. In particular, it should adequately consider the streaming nature of data and key related aspects, such as regional background, accurately capture both proximity and similarity while bridging the disparities, and effectively address the sparsity. However, these factors are often overlooked or difficult to incorporate. In this paper, we propose a novel streaming multi-granularity hierarchical spatio-temporal network. Initially, we innovate by incorporating remote sensing data, facilitating the creation of hierarchical multi-granularity structure and the comprehension of regional background. We construct multiple high-level risk prediction tasks to enhance model's ability to cope with sparsity. Subsequently, to capture and bridge spatial proximity and semantic similarity, region features and multi-view graph undergo encoding processes to distill effective representations, followed by a graph-enhanced representation alignment module that reconciles their disparities. At last, an alternating experience replay with a dual-memory buffer is employed to accommodate streaming data scenarios. Extensive experiments on two real datasets verify the superiority of our model against the state-of-the-art methods.

Sequential recommendation systems aim to predict the future behaviors of users based on their historical interactions. Despite the success of neural architectures like Transformer and Graph Neural Networks, these models often struggle with the inherent challenge of sparse data in accurately predicting future user behaviors. To alleviate the data sparsity problem, some methods leverage the contrastive learning to generate contrastive views, assuming the items appear discretely at the same time intervals and focusing on the sequence order. However, these approaches neglect the crucial temporal-aware collaborative patterns hidden within the user-item interactions, leading to a limited variety of contrastive pairs and less informative embeddings. The proposed framework,Temporal-awaregraph contrastive learning with theoretical guarantees for sequentialRecommendation (TagRec), integrates temporal-aware collaborative patterns with adaptive data augmentation to generate more informative user and item representations. TagRec employs a temporal-aware graph neural network to embed the original graph, then generates augmented graphs through the addition of interactions via latent user interest mining, the dropping of redundant interaction edges, and the perturbation of temporal information. Theoretical guarantees are provided that these augmentations enhance the graph’s utility. Extensive experiments on real-world datasets demonstrate the superiority of the proposed approach over the state-of-the-art recommendation methods.

Traffic accidents pose a significant risk to human health and property safety. Therefore, to prevent traffic accidents, predicting their risks has garnered growing interest. We argue that a desired prediction solution should demonstrate resilience to the complexity of traffic accidents. In particular, it should adequately consider the regional background, accurately capture both spatial proximity and semantic similarity, and effectively address the sparsity of traffic accidents. However, these factors are often overlooked or difficult to incorporate. In this paper, we propose a novel multi-granularity hierarchical spatio-temporal network. Initially, we innovate by incorporating remote sensing data, facilitating the creation of hierarchical multi-granularity structure and the comprehension of regional background. We construct multiple high-level risk prediction tasks to enhance model's ability to cope with sparsity. Subsequently, to capture both spatial proximity and semantic similarity, region feature and multi-view graph undergo encoding processes to distill effective representations. Additionally, we propose message passing and adaptive temporal attention module that bridges different granularities and dynamically captures time correlations inherent in traffic accident patterns. At last, a multivariate hierarchical loss function is devised considering the complexity of the prediction purpose. Extensive experiments on two real datasets verify the superiority of our model against the state-of-the-art methods.

The drug recommendation task aims to predict safe and effective drug prescriptions based on the patients' historical electronic health records (EHRs). However, existing drug recommendation models generally have two limitations. First, they neglect the inherent characteristics of multiple views existing in patients' clinical data (e.g., diagnoses and procedures), leading to fragmented and inconsistent patient representations. Second, they do not fully exploit drug label information. Most models do not explicitly establish a mapping relationship between drug labels and patients' historical visits. To address these two problems, we proposed a label-aware multi-view drug recommendation model named LAMRec. In particular, LAMRec uses a cross-attention module to fuse information from the diagnosis and procedure views, and increases the mutual information of patient multi-view representations through multi-view contrastive loss; the label-wise attention mechanism fully explores drug label information by constructing an adaptive mapping of drug-visit to generate personalized representations that are aware of the drug-related visit information. Experiments on three real world medical datasets demonstrated the superiority of LAMRec, with a relative reduction of 5.25% in DDI compared to the optimal baseline, a relative improvement of 4.20% in Jaccard similarity scores, and a relative improvement of 3.10% in F1 scores. We released the code online at: https://github.com/Tyunsen/LAMRec.

Graph neural networks (GNNs) have shown ad-vantages in graph-based analysis tasks. However, most existing methods have the homogeneity assumption and show poor performance on heterophilic graphs, where the linked nodes have dissimilar features and different class labels, and the semantically related nodes might be multi-hop away. To address this limitation, this paper presents GraphRARE, a general framework built upon node relative entropy and deep reinforcement learning, to strengthen the expressive capability of GNNs. An innovative node relative entropy, which considers node features and structural similarity, is used to measure mutual information between node pairs. In addition, to avoid the sub-optimal solutions caused by mixing useful information and noises of remote nodes, a deep reinforcement learning-based algorithm is developed to optimize the graph topology. This algorithm selects informative nodes and discards noisy nodes based on the defined node relative en-tropy. Extensive experiments are conducted on seven real-world datasets. The experimental results demonstrate the superiority of GraphRARE in node classification and its capability to optimize the original graph topology.

The next point-of-interest (POI) prediction is a significant task in location-based services, yet its complexity arises from the consolidation of spatial and semantic intent. This fusion is subject to the influences of historical preferences, prevailing location, and environmental factors, thereby posing significant challenges. In addition, the uneven POI distribution further complicates the next POI prediction procedure. To address these challenges, we enrich input features and propose an effective deep-learning method within a two-step prediction framework. Our method first incorporates remote sensing data, capturing pivotal environmental context to enhance input features regarding both location and semantics. Subsequently, we employ a region quad-tree structure to integrate urban remote sensing, road network, and POI distribution spaces, aiming to devise a more coherent graph representation method for urban spatial. Leveraging this method, we construct the QR-P graph for the user's historical trajectories to encapsulate historical travel knowledge, thereby augmenting input features with comprehensive spatial and semantic insights. We devise distinct embedding modules to encode these features and employ an attention mechanism to fuse diverse encodings. In the two-step prediction procedure, we initially identify potential spatial zones by predicting user-preferred tiles, followed by pinpointing specific POls of a designated type within the projected tiles. Empirical findings from four real-world location-based social network datasets underscore the remarkable superiority of our proposed approach over competitive baseline methods.

Query rewrite, which aims to improve query efficiency by altering an SQL query's structure without changing its result, has been an important research problem. In order to maintain equivalence between the rewritten query and the original one during rewriting, traditional query rewrite methods always rewrite the queries following certain rewrite rules. However, some problems still remain. First, existing methods of finding the optimal choice or sequence of rewrite rules are still limited and the process always costs a lot of resources. Methods involving discovering new rewrite rules typically require complicated proofs of structural logic or extensive user interactions. Second, current query rewrite methods usually rely highly on DBMS cost estimators which are often not accurate. In this paper, we address these problems by proposing a novel query rewrite method named LLM-R 2 , which leverages a large language model (LLM) to recommend rewrite rules for a database rewrite system. To further enhance the inference ability of the LLM in recommending rewrite rules, we train a contrastive model using a curriculum-based approach to learn query representations and select effective query demonstrations for the LLM. Experimental results show that our method significantly improves the query execution efficiency and outperforms the baseline methods. In addition, our method exhibits high robustness across different datasets.

Traffic speed histograms show the distribution of traffic speeds over a certain period. Traffic speed might not be recorded continuously, leading to missing histograms for some links on a road network. However, accurate imputation of missing histograms is a critical yet challenging task. This paper introduces a novel framework to address four previously unexplored dimensions crucial for precise traffic speed histogram estimation: regionality, proximity, sparsity, and volatility. First, to address the challenge of regionality and proximity, we employ a global partition graph that captures both regional and proximal correlations within the road network. Next, in response to the challenge of sparsity, the framework features a disentangled feature encoding pipeline, comprising a global encoder and a localized spatio-temporal encoder. This design allows for the effective handling of entangled spatio-temporal dimensions, thereby mitigating the issues related to input sparsity. In particular, the framework leverages graph neural networks and recurrent neural networks to capture spatial and temporal correlations. In addition, to encompass the complexities of spatio-temporal correlations both on global and local scales, we employ a two-layer fusion module with an attention-based mechanism for representation integration. Lastly, to mitigate the challenge of volatility due to missing values, we incorporate a self-supervised learning task using an auto-encoder framework, enhancing the stability and robustness of the encoding models. Extensive evaluations on two real-world datasets confirm that our method significantly outperforms state-of-the-art solutions in terms of both accuracy and robustness.

Knowledge tracing models based on deep learning can achieve impressive predictive performance by leveraging attention mechanisms. However, there still exist two challenges in attentive knowledge tracing (AKT): First, the mechanism of classical models of AKT demonstrates relatively low attention when processing exercise sequences with shifting knowledge concepts (KC), making it difficult to capture the comprehensive state of knowledge across sequences. Second, classical models do not consider stochastic behaviors, which negatively affects models of AKT in terms of capturing anomalous knowledge states. This article proposes a model of AKT, called Enhancing Locality for Attentive Knowledge Tracing (ELAKT), that is a variant of the deep KT model. The proposed model leverages the encoder module of the transformer to aggregate knowledge embedding generated by both exercises and responses over all timesteps. In addition, it uses causal convolutions to aggregate and smooth the states of local knowledge. The ELAKT model uses the states of comprehensive KCs to introduce a prediction correction module to forecast the future responses of students to deal with noise caused by stochastic behaviors. The results of experiments demonstrated that the ELAKT model consistently outperforms state-of-the-art baseline KT models.

With the rapid development of edge computing in the post-COVID19 pandemic period, precise workload forecasting is considered the basis for making full use of the edge limited resources, and both edge service providers (ESPs) and edge service consumers (ESCs) can benefit significantly from it. Existing paradigms of workload forecasting (i.e., edge-only or cloud-only) are improper, due to failing to consider the inter-site correlations and might suffer from significant data transmission delays. With the increasing adoption of edge platforms by web services, it is critical to balance both accuracy and efficiency in workload forecasting. In this paper, we propose ELASTIC, which is the first study that leverages a cloud-edge collaborative paradigm for edge workload forecasting with multi-view graphs. Specifically, at the global stage, we design a learnable aggregation layer on each edge site to reduce the time consumption while capturing the inter-site correlation. Additionally, at the local stage, we design a disaggregation layer combining both the intra-site correlation and inter-site correlation to improve the prediction accuracy. Extensive experiments on realistic edge workload datasets collected from China’s largest edge service provider show that ELASTIC outperforms state-of-the-art methods, decreases time consumption, and reduces communication cost.

Video streaming applications (VSAs) are increasingly being deployed on large-scale edge platforms, which have the potential to significantly improve the quality of service (QoS) and end-user experience (QoE), ultimately maximizing business outcomes. However, there is currently very little understanding of how QoS, QoE, and the impact of QoS on QoE for VSAs on edge platforms in the wild and at scale. To close the knowledge gap, we collect SNESet, an active measurement dataset comprising QoS and QoE telemetry metrics of 8 VSAs over four months, covering end-users from 798 edge sites,30 cities, and 3 ISPs in one country.We characterize and compare the QoS and QoE metrics in SNESet with existing publicly available datasets, highlighting that SNESet includes a significantly greater number of metrics (horizontal diversity and vertical hierarchy) and provides more comprehensive coverage of specific metrics.Moreover, we qualitatively and quantitatively analyze the impact of QoS on QoE in both domain-general and domain-specific scenarios. Our findings can inform the system design decisions that different entities in the video ecosystem (content providers, video player designers, third-party optimizers, edge vendors) make to maximize end-users experience and ultimately maximize the business outcomes. We hope SNESet can attract more research efforts in the data management community, computer network community, and beyond.

Extracting roads from multi-source data, such as aerial images and vehicle trajectories, is an important way to maintain road networks in the filed of urban computing. In this paper, we revisit the problem of road extraction and aim to boost its accuracy by solving three significant issues: the insufficient complementarity among multiple sources, rough edges of extracted roads, and many false positives caused by confusing pixels. In particular, we design an end-to-end neural network model to achieve this goal. At first, this model leverages two encoding networks to extract relative information from the inputs of two sources respectively, and then applies the attention mechanism to fuse them for sufficiently capturing the complementary correlation. Next, we introduce an auxiliary task, predicting road edges based on fused representations, to make the extracted roads smooth and continuous. At last, to reduce false positives relative to confusing pixels, we propose a pixel-aware contrastive-learning module to distinguish positive (roads) and negative (objects similar to roads) pixels. In addition, to improve the model's learning effectiveness, we propose a model-agnostic transfer learning method, which first builds auxiliary tasks to pre-train the whole model, and then fine-tunes the model's parameters for the main task. Extensive experiments on real datasets verify the superiority of our method as well as the importance of solving the three issues outlined above.

Materialized views (MVs) can significantly optimize the query processing in databases. However, it is hard to generate MVs for ordinary users because it relies on background knowledge, and existing methods rely on DBAs to generate and maintain MVs. However, DBAs cannot handle large-scale databases, especially cloud databases that have millions of database instances and support millions of users. Thus it calls for an autonomous MV management system. In this paper, we propose an autonomous materialized view management system. It analyzes query workloads, estimates the costs and benefits of materializing queries as views, and selects MVs to maximize the benefit within a space budget. We propose a serialization and encoding method that can encode the features of both queries and views into vectors. Then we design a sequence-to-sequence model, Encoder-Reducer, to estimate MVs' cost/benefit by taking the encoding vectors as input. Next, we propose a deep reinforcement learning model to select high-quality MVs, which enriches the state representation with Encoder-Reducer's output. Experimental results show that our method outperforms existing studies in terms of MV selection quality.

Estimated time of arrival (ETA) prediction, also known as travel time estimation, is a fundamental task for a wide range of intelligent transportation applications, such as navigation, route planning, and ride-hailing services. To accurately predict the travel time of a route, it is essential to take into account both contextual and predictive factors, such as spatial-temporal interaction, driving behavior, and traffic congestion propagation inference. The ETA prediction models previously deployed at Baidu Maps have addressed the factors of spatial-temporal interaction (ConSTGAT) and driving behavior (SSML). In this work, we believe that modeling traffic congestion propagation patterns is of great importance toward accurately performing ETA prediction, and we focus on this factor to improve ETA performance. Traffic congestion propagation pattern modeling is challenging, and it requires accounting for impact regions over time and cumulative effect of delay variations over time caused by traffic events on the road network. In this paper, we present a practical industrial-grade ETA prediction framework named DuETA. Specifically, we construct a congestion-sensitive graph based on the correlations of traffic patterns, and we develop a route-aware graph transformer to directly learn the long-distance correlations of the road segments. This design enables DuETA to capture the interactions between the road segment pairs that are spatially distant but highly correlated with traffic conditions. Extensive experiments are conducted on large-scale, real-world datasets collected from Baidu Maps. Experimental results show that ETA prediction can significantly benefit from the learned traffic congestion propagation patterns, which demonstrates the effectiveness and practical applicability of DuETA. In addition, DuETA has already been deployed in production at Baidu Maps, serving billions of requests every day. This demonstrates that DuETA is an industrial-grade and robust solution for large-scale ETA prediction services.

Entity linking aims to link ambiguous mentions to their corresponding entities in a knowledge base, which is significant and fundamental for various downstream applications, e.g., knowledge base completion, question answering, and information extraction. While great efforts have been devoted to this task, most of these studies follow the assumption that large-scale labeled data is available. However, when the labeled data is insufficient for specific domains due to labor-intensive annotation work, the performance of existing algorithms will suffer an intolerable decline. In this paper, we endeavor to solve the problem of few-shot entity linking, which only requires a minimal amount of in-domain labeled data and is more practical in real situations. Specifically, we firstly propose a novel weak supervision strategy to generate non-trivial synthetic entity-mention pairs based on mention rewriting. Since the quality of the synthetic data has a critical impact on effective model training, we further design a meta-learning mechanism to assign different weights to each synthetic entity-mention pair automatically. Through this way, we can profoundly exploit rich and precious semantic information to derive a well-trained entity linking model under the few-shot setting. The experiments on real-world datasets show that the proposed method can extensively improve the state-of-the-art few-shot entity linking model and achieve impressive performance when only a small amount of labeled data is available. Moreover, we also demonstrate the outstanding ability of the model's transferability. Our code and models will be open-sourced.

In this paper, we revisit the problem of route travel time estimation on a road network and aim to boost its accuracy by capturing and utilizing spatio-temporal features from four significant aspects: heterogeneity, proximity, periodicity and dynamicity. Spatial-wise, we consider two forms of heterogeneity at link level in a road network: the turning ways between different links are heterogeneous which can make the travel time of the same link various; different links contain heterogeneous attributes and thereby lead to different travel time. In addition, we take into account the proximity: neighboring links have similar traffic patterns and lead to similar travel speeds. To this end, we build a link-connection graph to capture such heterogeneity and proximity. Temporal-wise, the weekly/daily periodicity of temporal background information (e.g., rush hours) and dynamic traffic conditions have significant impact on the travel time, which result in static and dynamic spatio-temporal features respectively. To capture such impacts, we regard the travel time/speed as a combination of static and dynamic parts, and extract many spatio-temporal relevant features for the prediction task. Talking about the methodology, it remains an open problem to build a generic learning model to boost the estimation accuracy. Hence, we design a novel encoder-decoder framework - The encoder uses the sequence attention model to encode dynamic features from the temporal-wise perspective. The decoder first uses the heterogeneous graph attention model to decode the static part of travel speed based on static spatio-temporal features, and then leverages the sequence attention model to decode the estimated travel time from spatial-wise perspective. Extensive experiments on real datasets verify the superiority of our method as well as the importance of the four aspects outlined above.

Modern location-based systems have stimulated explosive growth of urban trajectory data and promoted many real-world applications, e.g. , trajectory prediction. However, heavy big data processing overhead and privacy concerns hinder trajectory acquisition and utilization. Inspired by regular trajectory distribution on transportation road networks, we propose to model trajectory data privately with a deep generative model and leverage the model to generate representative trajectories for downstream tasks or directly support these tasks ( e.g. , popularity ranking), rather than acquiring and processing the original big trajectory data. Nevertheless, it is rather challenging to model high-dimensional trajectories with time-varying yet skewed distribution. To address this problem, we model and generate trajectory sequence with judiciously encoded spatio-temporal features over skewed distribution by leveraging an important factor neglected by the literature - the underlying road properties ( e.g. , road types and directions), which are closely related to trajectory distribution. Specifically, we decompose trajectory into map-matched road sequence with temporal information and embed them to encode spatio-temporal features. Then, we enhance trajectory representation by encoding inherent route planning patterns from the underlying road properties. Later, we encode spatial correlations among edges and daily and weekly temporal periodicity information. Next, we employ a meta-learning module to generate trajectory sequence step by step by learning generalized trajectory distribution patterns from skewed trajectory data based on the well-encoded trajectory prefix. Last but not least, we preserve trajectory privacy by learning the model differential privately with clipping gradients. Experiments on real-world datasets show that our method significantly outperforms existing methods.

In this paper, we study how to jointly predict travel demands and traffic flows for all regions of a city at a future time interval. From an empirical analysis of traffic data, we outline three desired properties, namely region-level correlations, temporal periodicity and inter-traffic correlations. Then, we propose a comprehensive neural network based traffic prediction model, where various effective embeddings or encodings are designed to capture the aforementioned properties. First, we design effective region embeddings to capture two forms of region-level correlations: spatially close regions have similar embeddings, and regions with similar properties (e.g., the number of POIs and the number of roads in a region) other than locations have similar embeddings. Second, we extract the "day-in-week" and "time-in-day" and utilize the temporal periodicity in designing the embeddings for time intervals. Third, we propose an effective encoding for past traffic data which captures two forms of inter-traffic correlations - the correlation between past and future traffic, and the correlation between travel demands and traffic flows within past traffic data. Extensive experiments on two real datasets verify the high effectiveness of our model.

Materialized views (MVs) can significantly optimize the query processing in databases. However, it is hard to generate MVs for ordinary users because it relies on background knowledge, and existing methods rely on DBAs to generate and maintain MVs. However, DBAs cannot handle large-scale databases, especially cloud databases that have millions of database instances and support millions of users. Thus it calls for an autonomous MV management system. In this paper, we propose an autonomous materialized view management system, AutoView. It analyzes query workloads, estimates the costs and benefits of materializing queries as views, and selects MVs to maximize the benefit within a space budget. We propose a deep reinforcement learning model to select high-quality MVs, which enriches the state representation with query and MVs' embedding. Experimental results show that our method outperforms existing studies in terms of MV selection quality.

Abstract Intelligent transportation (e.g., intelligent traffic light) makes our travel more convenient and efficient. With the development of mobile Internet and position technologies, it is reasonable to collect spatio-temporal data and then leverage these data to achieve the goal of intelligent transportation, and here, traffic prediction plays an important role. In this paper, we provide a comprehensive survey on traffic prediction, which is from the spatio-temporal data layer to the intelligent transportation application layer. At first, we split the whole research scope into four parts from bottom to up, where the four parts are, respectively, spatio-temporal data, preprocessing, traffic prediction and traffic application. Later, we review existing work on the four parts. First, we summarize traffic data into five types according to their difference on spatial and temporal dimensions. Second, we focus on four significant data preprocessing techniques: map-matching, data cleaning, data storage and data compression. Third, we focus on three kinds of traffic prediction problems (i.e., classification, generation and estimation/forecasting). In particular, we summarize the challenges and discuss how existing methods address these challenges. Fourth, we list five typical traffic applications. Lastly, we provide emerging research challenges and opportunities. We believe that the survey can help the partitioners to understand existing traffic prediction problems and methods, which can further encourage them to solve their intelligent transportation applications.

Materializing views is an important method to reduce redundant computations in DBMS, especially for processing large scale analytical queries. However, many existing methods still need DBAs to manually generate materialized views, which are not scalable to a large number of database instances, especially on the cloud database. To address this problem, we propose an automatic view generation method which judiciously selects "highly beneficial" subqueries to generate materialized views. However, there are two challenges. (1) How to estimate the benefit of using a materialized view for a query? (2) How to select optimal subqueries to generate materialized views? To address the first challenge, we propose a neural network based method to estimate the benefit of using a materialized view to answer a query. In particular, we extract significant features from different perspectives and design effective encoding models to transform these features into hidden representations. To address the second challenge, we model this problem to an ILP (Integer Linear Programming) problem, which aims to maximize the utility by selecting optimal subqueries to materialize. We design an iterative optimization method to select subqueries to materialize. However, this method cannot guarantee the convergence of the solution. To address this issue, we model the iterative optimization process as an MDP (Markov Decision Process) and use the deep reinforcement learning model to solve the problem. Extensive experiments show that our method outperforms existing solutions by 28.4%, 8.8% and 31.7% on three real-world datasets.

In this paper, we study the problem of origin-destination (OD) travel time estimation where the OD input consists of an OD pair and a departure time. We propose a novel neural network based prediction model that fully exploits an important fact neglected by the literature -- for a past OD trip its travel time is usually affiliated with the trajectory it travels along, whereas it does not exist during prediction. At the training phase, our goal is to design novel representations for the OD input and its affiliated trajectory, such that they are close to each other in the latent space. First, we match the OD pairs and their affiliated (historical) trajectories to road networks, and utilize road segment embeddings to represent their spatial properties. Later, we match the timestamps associated with trajectories to time slots and utilize time slot embeddings to represent the temporal properties. Next, we build a temporal graph to capture the weekly and daily periodicity of time slot embeddings. Last, we design an effective encoding to represent the spatial and temporal properties of trajectories. To bind each OD input to its affiliated trajectory, we also encode the OD input into a hidden representation, and make the hidden representation close to the spatio-temporal representation of the trajectory. At the prediction phase, we only use the OD input, get the hidden representation of the OD input, and use it to generate the travel time. Extensive experiments on real datasets show that our method achieves high effectiveness and outperforms existing methods.

Many applications, e.g., Uber, collect large-scale trajectory data from moving vehicles on road network. Trajectory data analytics can benefit many real-world applications, such as route planning and transportation optimizations. Two core operations in trajectory data analytics are trajectory similarity search and join, and both of them rely on a trajectory similarity function to measure the similarity between two trajectories. However, existing similarity functions focus on trajectory points distance and neglect the fact the trajectories should be on road network. Obviously aligning trajectories on road network can remove the noise points introduced by system errors. Toward this goal, we define a road-network-aware trajectory similarity function to measure trajectory similarity. To support trajectory similarity search and join, we propose a filtering-refine framework. In the filtering step, we compute a signature of each trajectory such that if two trajectories are similar, they must share a common signature. We utilize the signatures to prune a huge number of dissimilar pairs. In the refine step, we design effective algorithms to verify the candidates that are not pruned in the filtering step. To support large-scale trajectories, we develop a system DISON for Distributed In-Memory Trajectory Similarity Search and Join on Road Network. DISON splits trajectories into disjoint partitions by considering load balance and locality, and designs effective global index to prune irrelevant partitions. Extensive experiments on real datasets showed that our method achieved high effectiveness, efficiency, and scalability and outperformed existing solutions significantly.

Trajectory data analytics plays an important role in many applications, such as transportation optimization, urban planning, taxi scheduling, and so on. However, trajectory data analytics has a great challenge that the time cost for processing queries is too high on big datasets. In this paper, we demonstrate a distributed in-memory framework Ratel base on Spark for analyzing large scale trajectories. Ratel groups trajectories into partitions by considering the data locality and load balance. We build R-Tree based global indexes to prune partitions when applying trajectory search and join. For each partition, Ratel uses a filter-refinement method to efficiently find similar trajectories. We show three kinds of scenarios - bus station planning, route recommendation, and transportation analytics. Demo attendees can interact with a web UI, pose different queries on the dataset, and navigate the query result.

Knowledge graphs are widely applied in many applications. Automatically solving mathematical exercises is also an interesting task which can be enhanced by knowledge reasoning. In this paper, we design MathGraph, a knowledge graph aiming to solve high school mathematical exercises. Since it requires fine-grained mathematical derivation and calculation of different mathematical objects, we design a crowdsourcing-based method to help build MathGraph. MathGraph supports massive kinds of mathematical objects, operations and constraints which may be involved in exercises. Furthermore, we propose an algorithm to align a semantically parsed exercise to MathGraph and figure out the answer automatically. Extensive experiments on real-world datasets verify the effectiveness of MathGraph.

Crowd-powered database systems can leverage the crowd's ability to address machine-hard problems, e.g., data integration. Existing crowdsourcing systems adopt the traditional tree model to select a good query plan. However, the tree model can optimize the I/O cost but cannot optimize the monetary cost, latency and quality, which are three important optimization goals in crowdsourcing. To address this limitation, we demonstrate CDB, a crowd-powered database system. CDB proposes a new graph-based model that adopts a fine-grained tuple-level optimization model which significantly outperforms existing coarse-grained tree-based optimization models. Moreover, CDB provides a unified framework to simultaneously optimize the monetary cost, quality and latency. We have deployed CDB on well-known crowd-sourcing platforms and users can easily use our system to deploy their applications. We will demonstrate how to use CDB to address real-world applications, including web table integration and entity collection.

Crowdsourcing database systems have been proposed to leverage crowd-powered operations to encapsulate the complexities of interacting with the crowd. Existing systems suffer from two major limitations. Firstly, in order to optimize a query, they often adopt the traditional tree model to select an optimized table-level join order. However, the tree model provides a coarse-grained optimization, which generates the same order for different joined tuples and limits the optimization potential that different joined tuples can be optimized by different orders. Secondly, they mainly focus on optimizing the monetary cost. In fact, there are three optimization goals (i.e., smaller monetary cost, lower latency, and higher quality) in crowdsourcing, and it calls for a system to enable multi-goal optimization.