Multivariate time series forecasting is fundamental to web-scale systems. However, frequency-domain forecasters face two structural challenges: (i) frequency bias, where low-amplitude yet informative temporal cues are often submerged in noise and neglected during training; (ii) spectral degradation, where standard neural transformations distort high-amplitude periodic structures, thereby weakening the predictive signal. To address both issues, we propose GFMixer, a decoupled dual-path architecture. GFMixer first applies a Temporal Gradient Block (TGB) to capture low-amplitude information through adaptive selection based on temporal-geometric evidence. Subsequently, a Fourier-Aware Attention Block (FAB) represents high-amplitude, multi-frequency information to mitigate spectral degradation. Finally, both information streams are integrated via time-aligned residual connections within a Gradient Aggregation Block (GAB) for the final forecasting task. Extensive evaluations across seven standard benchmarks (ETT, Weather, Electricity, and Traffic) demonstrate GFMixer's superiority, securing 37 first-place rankings and the lowest average MSE overall. Beyond standalone performance, GFMixer serves as a versatile plug-and-play module that consistently enhances mainstream backbones. Code is available at: https://github.com/superlin30/GFMixer.

In many real-world systems, the evolution of one time series often leads or lags that of its related peers rather than moving in perfect synchrony. Graph Neural Networks (GNNs) are widely used to model such inter-Series Lead-Lag relationships, representing entities as graph nodes with time-series attributes. However, existing methods typically collapse temporal information into discrete points and adopt uniform messaging mechanisms assuming synchronized upward/downward effects and identical time lags among related peers, which are often inconsistent with real-world dynamics. Furthermore, stacking GNN layers to capture multi-hop influences reduces interpretability, hindering understanding of underlying dynamics. To address these issues, we propose the Lead-LagNet, a framework designed to capture diverse cross-series propagation patterns with lead-lag phenomenon in time series. The Lead-LagNet identifies meaningful subsequences in time series and employs a gating mechanism to establish lead-lag connections, enabling the model to uncover complex influencing patterns without relying on predefined relationships. By decoupling the linear messaging process from non-linear feature extraction, the proposed Lead-LagNet enhances both modeling flexibility and interoperability. Experimental evaluation of both synthetic tasks and real-world datasets demonstrates the superiority of Lead-LagNet over state-of-the-art algorithms, including BiGRU, SFM, TGC, FinGAT and ADGAT. Our code and data are available at https://github.com/FICLAB/LeadLagNet.

Recent advancements in time-series forecasting have highlighted the importance of frequency-domain modeling. However, deep learning models primarily operate in the time domain, limiting their ability to capture frequency-based patterns. Existing approaches normally introduce novel neural network architectures tailored to task-specific frequency properties, yet they often lack generalization and require extensive domain-specific adaptations. In this paper, we propose FAT, a novel pretraining framework that learns generalizable Frequency-Aware Time-series representations through self-supervised learning. The key idea of FAT is to pretrain any backbone model to directly extract generalizable frequency patterns from time-domain signals and encode them into robust representations-eliminating the need for architectural modifications or additional modules during inference. This is achieved via a frequency reformer that amplifies critical frequency components learned through self-supervision and enforces similarity between the original and frequency-reformed time-series representations produced by the encoder. In addition, recognizing that semantically equivalent time-series can exhibit different frequency expressions-analogous to how the same phrase is pronounced differently by different speakers-FAT introduces a Knowledge-Guided Frequency Reformer that unifies the expression of frequency patterns with the same underlying semantics and extends similarity constraints to frequency-invariant augmented samples to enhance robustness of learned representation. Experiments on 14 benchmark datasets across regression and classification tasks show that FAT consistently achieves state-of-the-art performance while maintaining robustness across diverse backbone models, significantly outperforming existing pretraining methods. Our code is available at https://github.com/JiaXiangfei/FAT.

In the realm of deep graph learning, our study uniquely addresses the under-explored area of vague graph learning. While the effectiveness of deep graph learning is recognized across various disciplines, the nuances of vague graph learning — whether its inherent vagueness should be incorporated or disregarded and its influence on deep graph learning efficiency — remain largely unexamined. We fill this gap by introducing a novel decoupled graph learning framework. This is achieved by proposing a matrix-based or a tensor-based fusion module to estimate unobservable node attributes, a hybrid attention mechanism to bridge nodes with both explicit and implicit relationships, and a message-passing mechanism for feature-sensitive transporting. The design principle of decoupling allows it to accommodate ambiguities in any or all of these aspects of node representation, linking, and message passing. Furthermore, we leverage an extensive stock dataset spanning 64 years across the entire U.S. market to assess our framework. This real-world data not only adds a practical dimension to our study but also highlights the effectiveness of vague graph learning. Remarkably, our framework demonstrates superiority over state-of-the-art algorithms, marking performance enhancements of at least 6.73%, 7.25%, and 11.39% in terms of Rank IC,$R^{2}$, and Rank ICIR, respectively.

Stock Movement Prediction (SMP) aims at predicting listed companies' stock future price trend, which is a challenging task due to the volatile nature of financial markets. Recent financial studies show that the momentum spillover effect plays a significant role in stock fluctuation. However, previous studies typically only learn the simple connection information among related companies, which inevitably fail to model complex relations of listed companies in real financial market. To address this issue, we first construct a more comprehensive Market Knowledge Graph (MKG) which contains bi-typed entities including listed companies and their associated executives, and hybrid-relations including the explicit relations and implicit relations. Afterward, we proposeDanSmp, a novel Dual Attention Networks to learn the momentum spillover signals based upon the constructed MKG for stock prediction. The empirical experiments on our constructed datasets against nine SOTA baselines demonstrate that the proposedDanSmpis capable of improving stock prediction with the constructed MKG.

Bi-typed multi-relational heterogeneous graph (BMHG) is one of the most common graphs in practice, for example, academic networks, e-commerce user behavior graph and enterprise knowledge graph. It is a critical and challenge problem on how to learn the numerical representation for each node to characterize subtle structures. However, most previous studies treat all node relations in BMHG as the same class of relation without distinguishing the different characteristics between the intra-type relations and inter-type relations of the bi-typed nodes, causing the loss of significant structure information. To address this issue, we propose a novelDualHierarchicalAttentionNetworks (DHAN) based on the bi-typed multi-relational heterogeneous graphs to learn comprehensive node representations with the intra-type and inter-type attention-based encoder under a hierarchical mechanism. Specifically, the former encoder aggregates information from the same type of nodes, while the latter aggregates node representations from its different types of neighbors. Moreover, to sufficiently model node multi-relational information in BMHG, we adopt a newly proposed hierarchical mechanism. By doing so, the proposed dual hierarchical attention operations enable our model to fully capture the complex structures of the bi-typed multi-relational heterogeneous graphs. Experimental results on various tasks against the state-of-the-arts sufficiently confirm the capability of DHAN in learning node representations on the BMHGs.

Knowledge graph entity typing (KGET) aims to infer missing entity typing instances in KGs, which is a significant subtask of KG completion. Despite of its progress, however, it still faces two non-trivial challenges: (i) most existing KGET methods extract features by encoding the existing entity typing tuples, while ignoring rich relational knowledge. (ii) they typically treat each entity typing tuple in KGs independently, and thus inevitably fail to take account of the inherent and valuable neighborhood information surrounding a tuple. To address these challenges, we build a novel Heterogeneous Relational Graph (HRG), and propose a Multiplex Relational Graph Attention Networks (MRGAT) to learn on HRG, and then utilize a Connecting Embeddings model (ConnectE) to make entity type inference. Specifically, the framework contains three components. Firstly, to effectively integrate the entity typing tuples and entity relation triples in KGs, we construct a HRG that consists of three semantic subgraphs. Secondly, we employ MRGAT to learn embeddings on HRG. In MRGAT, each subgraph of HRG is fed to its corresponding model that is capable of capturing neighborhood information. Finally, given the learned embeddings, we make entity type prediction by ConnectE. Experimental results validate the superiority of our model against various state-of-the-art baselines.

Tradition financial studies on asset pricing focused on the economic indicators and media information of a stock. Recent financial studies found that the momentum spillovers of relevant firms are salient as well for measuring asset risk. However, previous studies on asset pricing via machine learning only relied on partial of these market information types. In this study, a deep learning framework is proposed to combine these three market information types with different data structures, that is, numerical economic indicators represented as scalars, media represented as textual vectors, and the influences of related firms captured by graphs. More importantly, the unique data characteristics brought by such data fusion are well addressed in the proposed learning framework. Specifically, a matrix-based module is first proposed to fuse numerical economic data and textual media, which specifically considers the interactions of the fused features. Such fused information, along with the firm relevance represented in graphs, is further integrated by a novel self-adaptive graph neural network that can address the dynamic merging of multilinked listed firms. Experiments performed on real market data demonstrate the effectiveness of the proposed approach over state-of-the-art algorithms, including eLSTM, RGCN, and TGC.

Media-aware stock movements are well acknowledged by the behavioral finance. As the soul of a firm, CEO’s media behavior is critical to the operation of a firm. CEO’s exposure could have captured the investors’ attention and enhanced the media effect in the stock market in terms of the “eyeball economics”, or CEO’s overexposure could have attracted more attention than firm-specific news, which attenuate the media effect in the stock market due to the investors’ limited attention. This study systematically explores the role and the moderating effect of CEO’s media behavior on the relationship between media content and stock markets. Using daily frequency data for a sample of Chinese stocks, this study shows that higher CEO media exposure attenuates the media effect on stock markets, especially consumer-related stocks.

In finance, it is believed that market information, namely, fundamentals and news information, affects stock movements. Such media-aware stock movements essentially comprise a multimodal problem. Two unique challenges arise in processing these multimodal data. First, information from one data mode will interact with information from other data modes. A common strategy is to concatenate various data modes into one compound vector; however, this strategy ignores the interactions among different modes. The second challenge is the heterogeneity of the data in terms of sampling time. Specifically, fundamental data consist of continuous values sampled at fixed time intervals, whereas news information emerges randomly. This heterogeneity can cause valuable information to be partially missing or can distort the feature spaces. In addition, the study of media-aware stock movements in previous work has focused on the one-to-one problem, in which it is assumed that news affects only the performance of the stocks mentioned in the reports. However, news articles also impact related stocks and cause stock co-movements. In this article, we propose a tensor-based event-driven LSTM model to address these challenges. Experiments performed on the China securities market demonstrate the superiority of the proposed approach over state-of-the-art algorithms, including AZFinText, eMAQT, and TeSIA.

Stock market volatility is influenced by information release, dissemination, and public acceptance. With the increasing volume and speed of social media, the effects of Web information on stock markets are becoming increasingly salient. However, studies of the effects of Web media on stock markets lack both depth and breadth due to the challenges in automatically acquiring and analyzing massive amounts of relevant information. In this study, we systematically reviewed 229 research articles on quantifying the interplay between Web media and stock markets from the fields of Finance, Management Information Systems, and Computer Science. In particular, we first categorized the representative works in terms of media type and then summarized the core techniques for converting textual information into machine-friendly forms. Finally, we compared the analysis models used to capture the hidden relationships between Web media and stock movements. Our goal is to clarify current cutting-edge research and its possible future directions to fully understand the mechanisms of Web information percolation and its impact on stock markets from the perspectives of investors cognitive behaviors, corporate governance, and stock market regulation.

To study the influence of information on the behavior of stock markets, a common strategy in previous studies has been to concatenate the features of various information sources into one compound feature vector, a procedure that makes it more difficult to distinguish the effects of different information sources. We maintain that capturing the intrinsic relations among multiple information sources is important for predicting stock trends. The challenge lies in modeling the complex space of various sources and types of information and studying the effects of this information on stock market behavior. For this purpose, we introduce a tensor-based information framework to predict stock movements. Specifically, our framework models the complex investor information environment with tensors. A global dimensionality-reduction algorithm is used to capture the links among various information sources in a tensor, and a sequence of tensors is used to represent information gathered over time. Finally, a tensor-based predictive model to forecast stock movements, which is in essence a high-order tensor regression learning problem, is presented. Experiments performed on an entire year of data for China Securities Index stocks demonstrate that a trading system based on our framework outperforms the classic Top- N trading strategy and two state-of-the-art media-aware trading algorithms.

Reading and Commenting online news is becoming a common user behavior in social media. Discussion in the form of comments following news postings can be effectively facilitated if the service provider can recommend articles based on not only the original news itself but also the thread of changing comments. This turns the traditional news recommendation to a "discussion moderator" that can intelligently assist online forums. In this work, we present a framework to recommend relevant information in the forum-based social media using user comments. When incorporating user comments, we consider structural and semantic information carried by them. Experiments indicate that our proposed solutions provide an effective recommendation service.

In this paper, we propose a decentralized collaborative filtering (CF) approach based on P2P overlay network for the autonomous agents' environment. Experiments show that our approach is more scalable than traditional centralized CF filtering systems and alleviates the sparsity problem in distributed CF.

Many collaborative music recommender systems (CMRS) have succeeded in capturing the similarity among users or items based on ratings, however they have rarely considered about the available information from the multimedia such as genres, let alone audio features from the media stream. Such information is valuable and can be used to solve several problems in RS. In this paper, we design a CMRS based on audio features of the multimedia stream. In the CMRS, we provide recommendation service by our proposed method where a clustering technique is used to integrate the audio features of music into the collaborative filtering (CF) framework in hopes of achieving better performance. Experiments are carried out to demonstrate that our approach is feasible.

With the development of e-commerce and the proliferation of easily accessible information, recommender systems have become a popular technique to prune large information spaces so that users are directed toward those items that best meet their needs and preferences. While many collaborative recommender systems (CRS) have succeeded in capturing the similarity among users or items based on ratings to provide good recommendation, there are still some challenges for them to be a more efficient RS. In this paper, we address three problems in CRS (user bias, non-transitive association, and new item problem) and provide a new item-based probabilistic model approach in order to solve the addressed problems in hopes of achieving better performance. In this probabilistic model, items are classified into groups and predictions are made for users considering the Gaussian distribution of user ratings. Experiments on a real-word data set illustrate that our proposed approach is comparable with others.

Recommender system is a kind of Web intelligence techniques to make a daily information filtering for people. Clustering techniques have been applied to the item-based collaborative filtering framework to solve the cold start problem. It also suggests a way to integrate the content information into the collaborative filtering. Extensive experiments have been conducted on MovieLens data to analyze the characteristics of our technique. The results show that our approach contributes to the improvement of prediction quality of the item-based collaborative filtering, especially for the cold start problem.