Search

Numerous spatial-temporal learning methods have been proposed for crowd flow modeling, which is an important problem in Intelligent Transportation Systems (ITSs). However, most of the existing methods were designed to use data in one specific form to solve one particular task of crowd flow modeling and the shared patterns among different tasks have been largely ignored. In this paper, we investigate how to learn generic node representations that can simultaneously support various downstream tasks of crowd flow modeling. Along this line, we develop a continuous-time dynamic graph representation learning method based on Bootstrapping for Crowd Flow modeling (BootCF). Our approach follows a training procedure with two phases. In the pre-training phase, the continuous-time dynamic encoder converts edges with timestamps into messages to update the representations of the related traffic nodes. Inspired by the recent progress of contrastive learning, a bootstrapping framework for continuous-time dynamic graphs is designed to calculate pre-training loss and update the model in a self-supervised way, and thus enabling the node representation learning to be task-agnostic. Moreover, a context-aware data augmentation on continuous-time dynamic graphs is proposed to generate the augmented view of input data. Once the general node representations are obtained, the second phase can learn an effective model for any downstream task. Experiments on two real-world datasets show that our approach can achieve significant performance gain on four downstream tasks, which demonstrates that the proposed method has the powerful generalization capability for learning task-agnostic node representations.

Given a sequence of sets, where each set has a timestamp and contains an arbitrary number of elements, temporal sets prediction aims to predict the elements in the subsequent set. Previous studies for temporal sets prediction mainly focus on the modelling of elements and implicitly represent each user's preference based on his/her interacted elements. However, user preferences are often continuously evolving and the evolutionary trend cannot be fully captured with the indirect learning paradigm of user preferences. To this end, we propose a continuous-time user preference modelling framework for temporal sets prediction, which explicitly models the evolving preference of each user by maintaining a memory bank to store the states of all the users and elements. Specifically, we first construct a universal sequence by arranging all the user-set interactions in a non-descending temporal order, and then chronologically learn from each user-set interaction. For each interaction, we continuously update the memories of the related user and elements based on their currently encoded messages and past memories. Moreover, we present a personalized user behavior learning module to discover user-specific characteristics based on each user's historical sequence, which aggregates the previously interacted elements from dual perspectives according to the user and elements. Finally, we develop a set-batch algorithm to improve the model efficiency, which can create time-consistent batches in advance and achieve 3.5× and 3.0× speedups in the training and evaluation process on average. Experiments on four real-world datasets demonstrate the superiority of our approach over state-of-the-arts under both transductive and inductive settings. The good interpretability of our method is also shown.

Representation learning on heterogeneous graphs aims to obtain meaningful node representations to facilitate various downstream tasks, such as node classification and link prediction. Existing heterogeneous graph learning methods are primarily developed by following the propagation mechanism of node representations. There are few efforts on studying the role of relations for improving the learning of more fine-grained node representations. Indeed, it is important to collaboratively learn the semantic representations of relations and discern node representations with respect to different relation types. To this end, in this paper, we propose a Relation-aware Heterogeneous Graph Neural Network, namely R-HGNN, to learn node representations on heterogeneous graphs at a fine-grained level by considering relation-aware characteristics. Specifically, a dedicated graph convolution component is first designed to learn unique node representations from each relation-specific graph separately. Then, a cross-relation message passing module is developed to improve the interactions of node representations across different relations. Also, the relation representations are learned in a layer-wise manner to capture relation semantics, which are used to guide the node representation learning process. Moreover, a semantic fusing module is presented to aggregate relation-aware node representations into a compact representation with the learned relation representations. Finally, we conduct extensive experiments on a variety of graph learning tasks, and experimental results demonstrate that our approach consistently outperforms existing methods among all the tasks.

Data mining research has been significantly motivated by and benefited from real-world applications in novel domains. This special issue was proposed and edited to draw attention to domain-driven data mining and disseminate research in foundations, frameworks, and applications for data-driven and actionable knowledge discovery. Along with this special issue, we also organized a related workshop to continue the previous efforts on promoting advances in domain-driven data mining. This editorial report will first summarize the selected papers in the special issue, then discuss various industrial trends in the context of the selected papers, and finally document the keynote talks presented by the workshop. Although many scholars have made prominent contributions with the theme of domain-driven data mining, there are still various new research problems and challenges calling for more research investigations in the future. We hope this special issue is helpful for scholars working along this critically important line of research.

Real-world dynamic systems and complex objects are often monitored with multivariate time series where each dimension represents a system signal. Performing accurate diagnostic for a group of dynamic systems while simultaneously taking into account their similarities/distinctions, is a non-trivial task. In this paper, we develop an adaptive regularization approach to learning sparse connectivity structures in complex dynamic systems. The learned connectivity networks shed lights on the structural compositions of the system and hence can serve as highly informative inputs for various machine learning tasks. In particular, we focus on high-dimensional and semi-supervised learning scenarios and present a joint learning method to recover system-wise connectivity patterns by adaptively constructing a shared, sparsity-inducing regularization template across all systems. The shared template can be intuitively interpreted and used as a modeling template for efficient analysis of new systems. Moreover, our approach can flexibly incorporate information such as must-links and cannot-links for constructing regularization templates. Overall, our approach, named sparse adaptive regularization (SAR), can extract structure-related connectivity features efficiently and effectively, and result in significant improvements for machine learning tasks in dynamic systems. We benchmark our approach against the state-of-the-art methods with real-world data. Our results demonstrate the superiority of our approach over the baselines in terms of accuracy, efficiency, and interpretability.

The authors investigate how peer influence affects customers' product adoption behaviors in emerging video game platforms. Understanding peer influence is critical to motivating users' willingness to purchase and improving game publishers' marketing performance. While similarities between socially linked users can be viewed as a consequence of social influence, homophily may also contribute to such phenomenon, causing identification difficulties in observational studies. Using data from the world's largest digital distribution platform for video games, the authors leverage state-of-the-art recommender system algorithms and propose an innovative framework to identify social influence in the adoption of video games when a confounding homophily effect is present. The results show that peer influence has a positive impact on platform users' adoption behaviors (i.e., a user tends to adopt a video game that has been purchased by his peers). This study also finds that peer influence would have been overestimated if homophily was not properly controlled.

Data mining research has been significantly motivated by and benefited from real-world applications in novel domains. This special issue was proposed and edited to draw attention to domain-driven data mining and disseminate research in foundations, frameworks, and applications for data-driven and actionable knowledge discovery. Along with this special issue, we also organized a related workshop to continue the previous efforts on promoting advances in domain-driven data mining. This editorial report will first summarize the selected papers in the special issue, then discuss various industrial trends in the context of the selected papers, and finally document the keynote talks presented by the workshop. Although many scholars have made prominent contributions with the theme of domain-driven data mining, there are still various new research problems and challenges calling for more research investigations in the future. We hope this special issue is helpful for scholars working along this critically important line of research.

This article describes the use of tetramethylammonium fluoride tetrahydrate (TMAF·4H2O) for the large-scale preparation of a challenging 4-fluorothiazole. Commercially available TMAF·4H2O was procured on a large scale and rigorously dried by distillation with isopropyl alcohol and then dimethylformamide at elevated temperature. This method of drying provided anhydrous TMAF [TMAF (anh)] containing <0.2 wt % water and <60 ppm isopropanol. The use of TMAF (anh) was essential for production of the 4-fluorothiazole. When the chlorothiazole starting material was treated with other anhydrous fluoride sources, poor conversion of the starting material or potential safety issues were observed. SNAr fluorination using dried TMAF·4H2O was carried out at a 45.1 kg scale at 95–100 °C to produce 36.8 kg of 4-fluorothiazole 1b.

The synthesis of the highly energetic pharmaceutical intermediate 4-methyl-1,2,5-oxadiazole-3-carboxylic acid via cyclization of (2E,3E)-butane-2,3-dione dioxime followed by selective oxidation of the resulting 3,4-dimethyl-1,2,5-oxadiazole was developed to support the sample production for clinical trials. Rigorous process safety evaluations revealed many potential safety risks associated with the involved highly energetic compounds and the high exothermicity of the oxidation reaction. A continuous flow process was developed for the synthesis of 3,4-dimethyl-1,2,5-oxadiazole to mitigate the potential safety risks associated with the thermal instability of starting material (2E,3E)-butane-2,3-dione dioxime and product 3,4-dimethyl-1,2,5-oxadiazole. A much safer process for the highly exothermic oxidation of 3,4-dimethyl-1,2,5-oxadiazole involving portion-wise addition of potassium permanganate was developed to avoid accumulation of reactive chemicals. The desired product was isolated as N-methyl morpholine 4-methyl-1,2,5-oxadiazole-3-carboxylate to avoid handling of potentially explosive 4-methyl-1,2,5-oxadiazole-3-carboxylic acid in its solid form. The developed process was successfully scaled up to afford a total of 93.52 kg of N-methyl morpholine 4-methyl-1,2,5-oxadiazole-3-carboxylate to support the production of a drug substance.

There has been increasing interest in exploring the impact of human behavior on financial market dynamics. One of the important related questions is whether attention from society can lead to significant stock price movements or even abnormal returns. To answer the question, we develop a new measurement of social attention, named periodic cumulative degree of social attention, by simultaneously considering the individual influence and the information propagation in social networks. Based on the vast social network data, we evaluate the new attention measurement by testing its significance in explaining future abnormal returns. In addition, we test the forecasting ability of social attention for stock price shocks, defined by the cumulative abnormal returns. Our results provide significant evidence to support the intercorrelated relationship between the social attention and future abnormal returns. The outperformance of the new approach in predicting price shocks is also confirmed by comparison with several benchmark methods.