Existing research on Temporal Knowledge Graph (TKG) representation learning focuses on decomposing events into entities and relations, then employing various approaches to learn entity and relation representations in a low-dimensional vector space. However, all existing research overlooks the correlations among events, even though events are the core constituent elements of TKGs and involve heterogeneous correlations. To this end, we propose a Heterogeneous evolutionary Event cluster Aware Representation learning approach in Temporal Knowledge Graphs (HEART), which is the first event-centric approach in TKGs. Specifically, a heterogeneous event graph construction module is proposed to capture the diverse pairwise correlations between events by building co-entity and proximity heterogeneous edges. In addition, an event-aware multi-step evolutionary clustering module is proposed to capture continuous high-order correlations among events at different timestamps. Furthermore, an event cluster-aware unsupervised alignment mechanism is proposed to preserve the temporal smoothness of event clusters through cross-temporal alignment. Moreover, an event cluster-based self-supervised optimization mechanism is proposed to optimize the representations of event clusters. Experimental results on seven real-world datasets demonstrate that HEART achieves the state-of-the-art performance, outperforming the runner-up by an average of 3.89%, 8.14%, 5.97%, and 6.92% in MRR, Hits@1, Hits@3, and Hits@10, respectively.

Feature interaction modeling, which exploits interactive information between features, has been widely explored in various applications. Recently, many graph or hypergraph structures-based models have been proposed to model feature interactions by predicting the existence of edges or hyperedges among nodes. However, these models lack the capability to capture the inherent comparability among samples, where multiple samples exhibit both shared and distinct characteristics, and such comparable relationships are often beneficial for prediction. To this end, we propose SR-HyperFM, Sample Relationship aware Hypergraph Factorization Machines, which incorporate sample comparable relationships into feature interaction modeling, leveraging both shared features and critical differences among samples. Specifically, the sample relationship aware hypergraph construction module is introduced to fully capture the comparable relationships among samples and discover beneficial high-order feature interactions. In addition, the dual hypergraph message passing module explicitly models feature interactions by exploiting these inherent relationships. Extensive experiments on four real-world datasets demonstrate the superiority of SR-HyperFM. In addition, case studies are conducted to further justify the effectiveness of SR-HyperFM.

In multivariate time series (MTS) forecasting, many deep learning based methods have been proposed for modeling dependencies at multiple spatial (inter-variate) or temporal (intra-variate) scales. However, existing methods may fail to model dependencies across multiple spatial-temporal scales (ST-scales, i.e., scales that jointly consider spatial and temporal scopes). In this work, we propose ST-Hyper to model the high-order dependencies across multiple ST-scales through adaptive hypergraph modeling. Specifically, we introduce a Spatial-Temporal Pyramid Modeling (STPM) module to extract features at multiple ST-scales. Furthermore, we introduce an Adaptive Hypergraph Modeling (AHM) module that learns a sparse hypergraph to capture robust high-order dependencies among features. In addition, we interact with these features through tri-phase hypergraph propagation, which can comprehensively capture multi-scale spatial-temporal dynamics. Experimental results on six real-world MTS datasets demonstrate that ST-Hyper achieves the state-of-the-art performance, outperforming the best baselines with an average MAE reduction of 3.8% and 6.8% for long-term and short-term forecasting, respectively. Code is available at https://anonymous.4open.science/ST-Hyper-83E7.

Multi-variate time series (MTS) forecasting is crucial for various applications. Existing methods have shown promising results owing to their strong ability to capture intra- and inter-variate dependencies. However, these methods often overlook lead-lag dependencies at multiple grouping scales, failing to capture hierarchical lead-lag effects in complex systems. To this end, we propose MillGNN, a novel graph neural network-based method that learns multiple grouping scale lead-lag dependencies for MTS forecasting, which can comprehensively capture lead-lag effects considering variate-wise and group-wise dynamics and decays. Specifically, MillGNN introduces two key innovations: (1) a scale-specific lead-lag graph learning module that integrates cross-correlation coefficients and dynamic decaying features derived from real-time inputs and time lags to learn lead-lag dependencies for each scale, which can model evolving lead-lag dependencies with statistical interpretability and data-driven flexibility; (2) a hierarchical lead-lag message passing module that passes lead-lag messages at multiple grouping scales in a structured way to simultaneously propagate intra- and inter-scale lead-lag effects, which can capture multi-scale lead-lag effects with a balance of comprehensiveness and efficiency. Experimental results on 11 datasets demonstrate the superiority of MillGNN for long-term and short-term MTS forecasting, compared with 16 state-of-the-art methods.

Multivariate time series (MTS) forecasting has attracted much attention in many intelligent applications. It is not a trivial task, as we need to consider both intra-variable dependencies and inter-variable dependencies. However, existing works are designed for specific scenarios and require much domain knowledge and expert efforts, which is difficult to transfer between different scenarios. In this article, we propose a scale-aware neural architecture search framework for MTS forecasting (SNAS4MTF). A multi-scale decomposition module transforms raw time series into multi-scale sub-series, which can preserve multi-scale temporal patterns. An adaptive graph learning module infers the different inter-variable dependencies under different time scales without any prior knowledge. For MTS forecasting, a search space is designed to capture both intra-variable dependencies and inter-variable dependencies at each time scale. The multi-scale decomposition, adaptive graph learning, and neural architecture search modules are jointly learned in an end-to-end framework. Extensive experiments on two real-world datasets demonstrate that SNAS4MTF achieves a promising performance compared with the state-of-the-art methods.

The problem of air pollution threatens public health. Air quality forecasting can provide the air quality index hours or even days later, which can help the public to prevent air pollution in advance. Previous works focus on citywide air quality forecasting and cannot solve nationwide city forecasting problems, whose difficulties lie in capturing the latent dependencies between geographically distant but highly correlated cities. In this article, we propose the group-aware graph neural network (GAGNN), a hierarchical model for nationwide city air quality forecasting. The model constructs a city graph and a city group graph to model the spatial and latent dependencies between cities, respectively. GAGNN introduces a differentiable grouping network to discover the latent dependencies among cities and generate city groups. Based on the generated city groups, a group correlation encoding module is introduced to learn the correlations between them, which can effectively capture the dependencies between city groups. After the graph construction, GAGNN implements message passing mechanism to model the dependencies between cities and city groups. The evaluation experiments on two real-world nationwide city air quality datasets, including the China dataset and the US dataset, indicate that our GAGNN outperforms existing forecasting models.

Temporal Knowledge Graph (TKG) representation learning embeds entities and event types into a continuous low-dimensional vector space by integrating the temporal information, which is essential for downstream tasks, e.g., event prediction and question answering. Existing methods stack multiple graph convolution layers to model the influence of distant entities, leading to the over-smoothing problem. To alleviate the problem, recent studies infuse reinforcement learning to obtain paths that contribute to modeling the influence of distant entities. However, due to the limited number of hops, these studies fail to capture the correlation between entities that are far apart and even unreachable. To this end, we propose GTRL, an entity Group-aware Temporal knowledge graph Representation Learning method. GTRL is the first work that incorporates the entity group modeling to capture the correlation between entities by stacking only a finite number of layers. Specifically, the entity group mapper is proposed to generate entity groups from entities in a learning way. Based on entity groups, the implicit correlation encoder is introduced to capture implicit correlations between any pairwise entity groups. In addition, the hierarchical GCNs are exploited to accomplish the message aggregation and representation updating on the entity group graph and the entity graph. Finally, GRUs are employed to capture the temporal dependency in TKGs. Extensive experiments on six real-world datasets demonstrate that GTRL achieves the state-of-the-art performances on the event prediction task, outperforming the best baseline by an average of 7.35%, 6.09%, 8.31%, and 11.21% in MRR, Hits@1, Hits@3, and Hits@10, respectively.

Event prediction is a vital and challenging task in temporal knowledge graphs (TKGs), which have played crucial roles in various applications. Recently, many graph neural networks based approaches are proposed to model the graph structure information in TKGs. However, these approaches only construct graphs based on quadruplets and model the pairwise correlation between entities, which fail to capture the high-order correlations among entities. To this end, we propose DHyper, a recurrent Dual Hypergraph neural network for event prediction in TKGs, which simultaneously models the influences of the high-order correlations among both entities and relations. Specifically, a dual hypergraph learning module is proposed to discover the high-order correlations among entities and among relations in a parameterized way. A dual hypergraph message passing network is introduced to perform the information aggregation and representation fusion on the entity hypergraph and the relation hypergraph. Extensive experiments on six real-world datasets demonstrate that DHyper achieves the state-of-the-art performances, outperforming the best baseline by an average of 13.09%, 4.26%, 17.60%, and 18.03% in MRR, Hits@1, Hits@3, and Hits@10, respectively.

Traffic forecasting is an essential problem in urban planning and computing. The complex dynamic spatial-temporal dependencies among traffic objects (e.g., sensors and road segments) have been calling for highly flexible models; unfortunately, sophisticated models may suffer from poor robustness especially in capturing the trend of the time series (1st-order derivatives with time), leading to unrealistic forecasts. To address the challenge of balancing dynamics and robustness, we propose TrendGCN, a new scheme that extends the flexibility of GCNs and the distribution-preserving capacity of generative and adversarial loss for handling sequential data with inherent statistical correlations. On the one hand, our model simultaneously incorporates spatial (node-wise) embeddings and temporal (time-wise) embeddings to account for heterogeneous space-and-time convolutions; on the other hand, it uses GAN structure to systematically evaluate statistical consistencies between the real and the predicted time series in terms of both the temporal trending and the complex spatial-temporal dependencies. Compared with traditional approaches that handle step-wise predictive errors independently, our approach can produce more realistic and robust forecasts. Experiments on six benchmark traffic forecasting datasets and theoretical analysis both demonstrate the superiority and the state-of-the-art performance of TrendGCN. Source code is available at https://github.com/juyongjiang/TrendGCN.

Geo-tagged photo-based tourist attraction recommendation can discover users’ travel preferences from their taken photos, so as to recommend suitable tourist attractions to them. However, existing visual content-based methods cannot fully exploit the user and tourist attraction information of photos to extract visual features, and do not differentiate the significance of different photos. In this article, we propose multi-level visual similarity-based personalized tourist attraction recommendation using geo-tagged photos (MEAL). MEAL utilizes the visual contents of photos and interaction behavior data to obtain the final embeddings of users and tourist attractions, which are then used to predict the visit probabilities. Specifically, by crossing the user and tourist attraction information of photos, we define four visual similarity levels and introduce a corresponding quintuplet loss to embed the visual contents of photos. In addition, to capture the significance of different photos, we exploit the self-attention mechanism to obtain the visual representations of users and tourist attractions. We conducted experiments on two datasets crawled from Flickr, and the experimental results proved the advantage of this method.

Multivariate time series (MTS) forecasting plays an important role in the automation and optimization of intelligent applications. It is a challenging task, as we need to consider both complex intra-variable dependencies and inter-variable dependencies. Existing works only learn temporal patterns with the help of single inter-variable dependencies. However, there are multi-scale temporal patterns in many real-world MTS. Single inter-variable dependencies make the model prefer to learn one type of prominent and shared temporal patterns. In this article, we propose a multi-scale adaptive graph neural network (MAGNN) to address the above issue. MAGNN exploits a multi-scale pyramid network to preserve the underlying temporal dependencies at different time scales. Since the inter-variable dependencies may be different under distinct time scales, an adaptive graph learning module is designed to infer the scale-specific inter-variable dependencies without pre-defined priors. Given the multi-scale feature representations and scale-specific inter-variable dependencies, a multi-scale temporal graph neural network is introduced to jointly model intra-variable dependencies and inter-variable dependencies. After that, we develop a scale-wise fusion module to effectively promote the collaboration across different time scales, and automatically capture the importance of contributed temporal patterns. Experiments on six real-world datasets demonstrate that MAGNN outperforms the state-of-the-art methods across various settings.

Outcome prediction aims to predict the future health condition of patients from Electronic Health Record (EHR) data. Because of the sequential characteristic of EHR data, recurrent neural network (RNN)-based outcome prediction methods have achieved state-of-the-art results. However, the major drawback of RNN-based outcome prediction methods is lack of interpretability, which would lead to trust issues. Aiming at this problem, this paper proposes interpretable outcome prediction model with hierarchical attention (IoHAN), an interpretable outcome prediction model by leveraging attention mechanism. The main novelty of IoHAN is that it can pinpoint the fine-grained influence on the final prediction result of each medical component by decomposing the attention weights hierarchically into hospital visits, medical variables, and interactions between medical variables. We evaluated IoHAN on MIMIC-III, a large real-world EHR data set. The experiment results demonstrate that IoHAN can achieve higher prediction accuracy than state-of-the-art outcome prediction models. In addition, the hierarchical decomposed attention weights can interpret the prediction results in a more natural and understandable way.

Predicting user positive response (e.g., purchases and clicks) probability is a critical task in Web applications. To identify predictive features from raw data, the state-of-the-art extreme deep factorization machine model (xDeepFM) introduces a new interaction network to leverage feature interactions at the vector-wise level explicitly. However, since each hidden layer in the interaction network is a collection of feature maps, it can be viewed essentially as an ensemble of different feature maps. In this case, only using a single objective to minimize the prediction loss may lead to overfitting and generate correlated errors. In this article, an ensemble diversity enhanced extreme deep factorization machine model (DexDeepFM) is proposed, which designs the ensemble diversity measure in each hidden layer and considers both ensemble diversity and prediction accuracy in the objective function. In addition, the attention mechanism is introduced to discriminate the importance of ensemble diversity measures with different feature interaction orders. Extensive experiments on three public real-world datasets are conducted to show the effectiveness of the proposed model.

Temporal knowledge graph (TKG) representation learning embeds relations and entities into a continuous low-dimensional vector space by incorporating temporal information. Latest studies mainly aim at learning entity representations by modeling entity interactions from the neighbor structure of the graph. However, the interactions of relations from the neighbor structure of the graph are neglected, which are also of significance for learning informative representations. In addition, there still lacks an effective historical relation encoder to model the multi-range temporal dependencies. In this article, we propose a d ual gr a ph c onvolution network based TKG representation learning method using h istorical rel a tions (DACHA). Specifically, we first construct the primal graph according to historical relations, as well as the edge graph by regarding historical relations as nodes. Then, we employ the dual graph convolution network to capture the interactions of both entities and historical relations from the neighbor structure of the graph. In addition, the temporal self-attentive historical relation encoder is proposed to explicitly model both local and global temporal dependencies. Extensive experiments on two event based TKG datasets demonstrate that DACHA achieves the state-of-the-art results.

The proliferation of digital cameras and the growing practice of online photo sharing using social media sites such as Flickr have resulted in huge volumes of geotagged photos available on the Web. Based on users' traveling preferences elicited from their travel experiences exposed on social media sites by sharing geotagged photos, we propose a new method for recommending tourist locations that are relevant to users (i.e., personalization) in the given context (i.e., context awareness). We obtain user-specific travel preferences from his/her travel history in one city and use these to recommend tourist locations in another city. Our technique is illustrated on a sample of publicly available Flickr dataset containing photos taken in various cities of China. Results show that our context-aware personalized method is able to predict tourists' preferences in a new or unknown city more precisely and generate better recommendations compared to other state-of-the-art landmark recommendation methods.

A place is a locale that is frequently visited by an individual user and carries important semantic meanings (e.g. home, work, etc.). Many location-aware applications will be greatly enhanced with the ability of the automatic discovery of personally semantic places. The discovery of a user's personally semantic places involves obtaining the physical locations and semantic meanings of these places. In this paper, we propose approaches to address both of the problems. For the physical place extraction problem, a hierarchical clustering algorithm is proposed to firstly extract visit points from the GPS trajectories, and then these visit points can be clustered to form physical places. For the semantic place recognition problem, Bayesian networks (encoding the temporal patterns in which the places are visited) are used in combination with a customized POI (i.e. place of interest) database (containing the spatial features of the places) to categorize the extracted physical places into pre-defined types. An extensive set of experiments have been conducted to demonstrate the effectiveness of the proposed approaches based on a dataset of real-world GPS trajectories.

An average white-collar worker deals with enormous amount of digital information on daily basis. Recently, there has been a growing interest to support their work. However, in order to be really supportive there is a need to know the current activity of the user at all times. In this paper we present a new technique that takes advantage of temporal aspects of user activity behavior to infer when it is most likely that an activity switch is occurring. We then describe "Activity Switch Detector" an interactive switch notification system embodying these ideas, and an extensive user study by ten participants to test the validity of the approach and present its results.

This paper presents a system where the personal route of a user is predicted using a probabilistic model built from the historical trajectory data . Route patterns are extracted from personal trajectory data using a novel mining algorithm, Continuous Route Pattern Mining (CRPM), which can tolerate different kinds of disturbance in trajectory data. Furthermore, a client–server architecture is employed which has the dual purpose of guaranteeing the privacy of personal data and greatly reducing the computational load on mobile devices. An evaluation using a corpus of trajectory data from 17 people demonstrates that CRPM can extract longer route patterns than current methods. Moreover, the average correct rate of one step prediction of our system is greater than 71%, and the average Levenshtein distance of continuous route prediction of our system is about 30% shorter than that of the Markov model based method.