Industrial anomaly detection, which relies on the analysis of industrial internet of things (IIoT) sensor data, is a critical element for guaranteeing the quality and safety of industrial manufacturing. Current solutions normally apply edge–cloud IIoT architecture. The edge side collects sensor data in the field, while the cloud side receives sensor data and analyzes anomalies to accomplish it. The more complete the data sent to the cloud side, the higher the anomaly-detection accuracy that can be achieved. However, it will be extremely expensive to collect all sensor data and transmit them to the cloud side due to the massive amounts and distributed deployments of IIoT sensors requiring expensive network traffics and computational capacities. Thus, it becomes a trade-off problem: “How to reduce data transmission under the premise of ensuring the accuracy of anomaly detection?”. To this end, the paper proposes a binary-convolution data-reduction network for edge–cloud IIoT anomaly detection. It collects raw sensor data and extracts their features at the edge side, and receives data features to discover anomalies at the cloud side. To implement this, a time-scalar binary feature encoder is proposed and deployed on the edge side, encoding raw data into time-series binary vectors. Then, a binary-convolution data-reduction network is presented at the edge side to extract data features that significantly reduce the data size without losing critical information. At last, a real-time anomaly detector based on hierarchical temporal memory (HTM) is established on the cloud side to identify anomalies. The proposed model is validated on the NAB dataset, and achieves 70.0, 64.6 and 74.0 on the three evaluation metrics of SP, RLFP and RLFN, while obtaining a reduction rate of 96.19%. Extensive experimental results demonstrate that the proposed method achieves new state-of-the-art results in anomaly detection with data reduction. The proposed method is also deployed on a real-world industrial project as a case study to prove the feasibility and effectiveness of the proposed method.

No abstract

Aspect-based sentiment analysis (ABSA) is a task in natural language processing (NLP) that involves predicting the sentiment polarity towards a specific aspect in text. Graph neural networks (GNNs) have been shown to be effective tools for sentiment analysis tasks, but current research often overlooks affective information in the text, leading to irrelevant information being learned for specific aspects. To address this issue, we propose a novel GNN model, MHAKE-GCN, which is based on the graph convolutional neural network (GCN) and multi-head attention (MHA). Our model incorporates external sentiment knowledge into the GCN and fully extracts semantic and syntactic information from a sentence using MHA. By adding weights to sentiment words associated with aspect words, our model can better learn sentiment expressions related to specific aspects. Our model was evaluated on four publicly benchmark datasets and compared against twelve other methods. The results of the experiments demonstrate the effectiveness of the proposed model for the task of aspect-based sentiment analysis.

No abstract