Adaptive knowledge transfer from pretrained source models to unlabeled target domains has emerged as a key paradigm in the Industrial Internet of Things. Such frameworks effectively mitigate distribution discrepancies between domains and ensure data privacy in cross-condition fault diagnosis. However, most existing methods require extensive parameter tuning for each source backbone, resulting in high computational costs, especially as the number of source domains or the complexity of models increases. Moreover, current methods typically learn the transferable weights of each source model at domain-level, inevitably causing instance-level bias. Thus, we propose a new idea of Specificity and Consistency Ensemble (SpeConE) for efficient multiple-source fault diagnosis, which avoids parameter tuning for each source backbone. SpeConE balances instance-specific and domain-consistent weights by jointly considering feature-feature and feature-output similarities. By fine-tuning the source bottlenecks guided by the specificity and consistency weights, SpeConE effectively achieves multi-source-free cross-domain adaptation. Additionally, a novel pseudo-label smoothing strategy is introduced to prevent overfitting in the target domain. Experiments show that, when adopting Swin Transformer, SpeConE outperforms similar methods (e.g., ∼+3% on single-device diagnosis and ∼+10% on cross-device diagnosis) while tuning only about 5% of the parameters.

Structural anomaly detection is essential for ensuring the safety, reliability, and longevity of building engineering. By identifying deviations from normal patterns, it enables early intervention and prevents potential failures. However, most existing methods rely on a single feature extracted from the time domain or the frequency domain. Time-domain features alone are insufficient to capture variations in the frequency components, while frequency-domain features fail to account for transient behaviours in the time domain. This limitation significantly reduces detection performance, particularly when dealing with nonlinear and non-stationary signals. To address the issues, this study proposes a new framework for anomaly detection using parallel time convolutional networks (TCN) and wavelet decomposition based convolutional neural networks (WD-CNN), termed PTWC. In this framework, time-frequency domain features are achieved by utilizing TCN to capture time-domain features and WD-CNN to capture frequency-domain features, distinguishing structural anomaly patterns. The proposed PTWC framework is validated on an actual three-story frame structure. Compared with ten baseline methods, the experimental results demonstrate that PTWC has high accuracy and achieves at least 7 % improvement in area under the curve (AUC) scores, thus confirming its superior performance in structural anomaly detection.

Effective issue management is critical for the success of open-source projects on GitHub. However, the platform currently lacks the capability to identify implicit links between issues, complicating the management process. In this study, we propose a machine learning-based approach to identify and classify these links. Preliminary experimental results demonstrate the effectiveness of our approach in identifying issue links, indicating its potential to enhance issue tracking in large-scale GitHub projects.

Developers of mobile applications (apps) can enhance their work efficiency by reusing suitable third-party libraries (TPLs). TPLs recommendation methods have been proposed to assist app developers in quickly finding useful TPLs, but the existing methods, such as those based on graph neural networks (GNN), have limitations in extracting high-order neighborhood information that can enhance the representation ability of nodes. To address this issue, we propose a novel hypergraph neural network method based on collaborative filtering, called High-order Collaborative Filtering (HCF). We first build two hypergraphs by fully exploiting the TPLs usage records in apps and then extracting the high-order neighborhood information from the hypergraphs. The neighborhood information extracted contains less noise compared to classic GNNs, thereby alleviating the problem of over-smoothing in GNN node representations. This advantage enables HCF to recommend more accurate and diverse TPLs for apps development. Extensive experiments on a real-world dataset demonstrate that HCF significantly outperforms the state-of-the-art methods in terms of recommendation accuracy and diversity.

GitHub has become a prominent platform for open source software development, facilitating collaboration and communication among a diverse group of contributors. Efficient issue tracking is a crucial aspect of managing projects on GitHub, and labels serve as one of the primary mechanisms for issue prioritization, while various other issue features are also utilized by issue handlers for the same purpose. However, in large projects, prioritizing issues remains a challenge, and the efficacy of using labels or other issue features for prioritization is not well understood. To address this knowledge gap, we conduct a comprehensive empirical study that investigates the role of labels in GitHub issue prioritization, examines the influence of various issue features on prioritization, and assesses the performance of different ranking algorithms based on these impactful features. Our study, conducted on a dataset comprising data from over 1.5 million issues across diverse GitHub projects, provides valuable insights for issue handling in open source platforms and offers guidance for future research in this domain. Specifically, the study reveals the limited effectiveness of labels in issue prioritization, highlights the significance of certain issue features in the prioritization process, and compares the performance of various ranking algorithms for issue prioritization to support issue handlers.

The conversational recommender systems (CRSs) have received extensive attention in recent years. However, most of the existing works focus on various deep learning models, which are largely limited by the requirement of large-scale human-annotated datasets. Such methods are not able to deal with the cold-start scenarios in industrial products. To alleviate the problem, we propose FORCE, a Framework Of Rule-based Conversational rEcommender system that helps developers to quickly build CRS bots by simple configuration. We conduct experiments on two datasets in different languages and domains to verify its effectiveness and usability.

Though recent end-to-end neural models have shown the promising progress on Conversational Recommender System (CRS), two key challenges still remain.First, the recommended items cannot be always incorporated into the generated replies precisely and appropriately.Second, only the items mentioned in the training corpus have a chance to be recommended in the conversation.To tackle these challenges, we introduce a novel framework called NTRD for recommender dialogue system that decouples the dialogue generation from the item recommendation.NTRD has two key components, i.e., response template generator and item selector.The former adopts an encoder-decoder model to generate a response template with slot locations tied to target items, while the latter fills in slot locations with the proper items using a sufficient attention mechanism.Our approach combines the strengths of both classical slot filling approaches (that are generally controllable) and modern neural NLG approaches (that are generally more natural and accurate).Extensive experiments on the benchmark RE-DIAL show our NTRD significantly outperforms the previous state-of-the-art methods.Besides, our approach has the unique advantage to produce novel items that do not appear in the training set of dialogue corpus.