Level Semantic Information Research Articles

Orchard bird song recognition based on multi-view multi-level contrastive learning

In the harvest season, orchards are frequently plagued by birds, and thus significant fruit pecking can adversely affect both the fruit quality and yield. Recognizing bird songs is crucial for preventing damage caused by orchard birds because it can provide the basis for subsequent bird repellent efforts. However, the extensive effort required to annotate sound samples poses a significant challenge for supervised deep learning. In this paper, we propose a self-supervised multi-view learning framework based on multi-level contrasting (MV-MLC) for bird song recognition, which utilizes both time and spectrogram views as inputs. This framework leverages MLC to automatically learn representations from unlabeled data and a multi-scale feature extraction (MSFE) backbone network is employed to capture the temporal features of bird songs at different scales. The time-spectrogram consistency task in MLC learning facilitates semantic-level information exchange across multi-views, while the hierarchical contrastive learning task captures granularity-level information, thereby resulting in more robust contextual representations. In addition, embedding the shuffle attention module in MSFE facilitates mining of the spatial and channel dependencies of bird song features to further enhance the representation of features by the multi-scale network. We conducted extensive experiments using our self-built 10-class bird song data set (Orchard-birds) and the publicly available Birdsdata and Powdermill data sets. The experimental results demonstrated that MV-MLC performed better than state-of-the-art self-supervised models. In particular, MV-MLC obtained outstanding performance even with a small proportion of labeled data. The recognition accuracies based on the Orchard-birds and Birdsdata data sets were 99.40% and 92.67%, respectively, with macro F1-scores of 99.40% and 92.61%.

A robust multi-scale feature extraction framework with dual memory module for multivariate time series anomaly detection

Although existing reconstruction-based multivariate time series anomaly detection (MTSAD) methods have shown advanced performance, most assume the training data is clean. When faced with noise or contamination in training data, they can also reconstruct the anomaly well, weakening the distinction between normal and anomaly. Some probabilistic generation-based methods have been used to address this issue because of their implicit robust structure to noise, but the training process and suppression of anomalous generalization are not stable. The recently proposed explicit method based on the memory module would also sacrifice the reconstruction effect of normal patterns, resulting in limited performance improvement. Moreover, most existing MTSAD methods use a single fixed-length window for input, which weakens their ability to extract long-term dependency. This paper proposes a robust multi-scale feature extraction framework with the dual memory module to comprehensively extract features fusing different levels of semantic information and lengths of temporal dependency. First, this paper designs consecutive neighboring windows as inputs to allow the model to extract local and long-term dependency information. Secondly, a dual memory-augmented encoder is proposed to extract global typical patterns and local common features. It ensures the reconstruction ability of normal data while suppressing the generalization of the anomaly. Finally, this paper proposes a multi-scale fusion module to fuse latent variables representing different levels of semantic information and uses the reconstructed latent variables to reconstruct samples for anomaly detection. Experimental results on five datasets from diverse domains show that the proposed method outperforms 16 typical baseline methods.

LR3S: A lightweight semantic segmentation model for road scenes based on improved DeepLabV3+

In the field of autonomous driving, driving systems need to understand and quickly respond to changes in road scenes, which makes it equally important to enhance the accuracy and real-time performance of semantic segmentation tasks in road scenes. This article proposes a lightweight road scene semantic segmentation model LR3S that integrates global contextual information based on the DeepLabV3+ framework. LR3S utilizes a lightweight GhostNetV2 network as the backbone to capture rich semantic information in images, and uses ASPP_eSE module to enhance the capture of multi-scale and detail level semantic information. In addition, a lightweight CARAFE upsampling operator is utilized to upsample feature maps, taking advantage of CARAFE’s large receptive field and low computational cost to prevent the loss of fine-grained features and ensure the integrity of semantic information. Experimental results demonstrate that LR3S achieves an MIoU of 74.47% on the Cityscapes dataset and obtains an MIoU of 76.01% on the PASCAL VOC 2012 dataset. Compared to baseline semantic segmentation models, LR3S significantly reduces the parameter amount while maintaining segmentation accuracy, achieving a good balance between model accuracy and real-time performance.

Hybrid Masked Image Modeling for 3D Medical Image Segmentation.

Masked image modeling (MIM) with transformer backbones has recently been exploited as a powerful self-supervised pre-training technique. The existing MIM methods adopt the strategy to mask random patches of the image and reconstruct the missing pixels, which only considers semantic information at a lower level, and causes a long pre-training time. This paper presents HybridMIM, a novel hybrid self-supervised learning method based on masked image modeling for 3D medical image segmentation. Specifically, we design a two-level masking hierarchy to specify which and how patches in sub-volumes are masked, effectively providing the constraints of higher level semantic information. Then we learn the semantic information of medical images at three levels, including: 1) partial region prediction to reconstruct key contents of the 3D image, which largely reduces the pre-training time burden (pixel-level); 2) patch-masking perception to learn the spatial relationship between the patches in each sub-volume (region-level); and 3) drop-out-based contrastive learning between samples within a mini-batch, which further improves the generalization ability of the framework (sample-level). The proposed framework is versatile to support both CNN and transformer as encoder backbones, and also enables to pre-train decoders for image segmentation. We conduct comprehensive experiments on five widely-used public medical image segmentation datasets, including BraTS2020, BTCV, MSD Liver, MSD Spleen, and BraTS2023. The experimental results show the clear superiority of HybridMIM against competing supervised methods, masked pre-training approaches, and other self-supervised methods, in terms of quantitative metrics, speed performance and qualitative observations. The codes of HybridMIM are available at https://github.com/ge-xing/HybridMIM.

A semantic-driven coupled network for infrared and visible image fusion

In order to be adapted to high-level vision tasks, several infrared and visible image fusion methods cascade with the downstream network to enhance the semantic information of fusion results. However, due to the feature-level heterogeneities between fusion and downstream tasks, these methods suffer from the loss of pixel-level information and incomplete reconstruction of semantic-level information. To further improve the performance of fusion images in high-level vision tasks, we propose a semantic-driven coupled network for infrared and visible image fusion, terms as SDCFusion. Firstly, to address feature heterogeneity, we couple the segmentation and fusion networks into a joint framework such that both networks share the multi-level cross-modality coupled features. Based on the joint optimization of dual tasks, a joint action between fusion and downstream tasks is formed to force the cross-modality coupled features modeled on both pixel domain and semantic domain. Subsequently, to guide the semantic information reconstruction, we cascade two networks to form the semantic-based driven action, which continuously optimizes the fusion image to achieve semantic representation capacity. In addition, we introduce an adaptive training strategy to reduce the complexity of dual-task training. Specifically, an mIoU-based semantic measurement weight is designed to balance the joint action and driven action throughout the training process. We evaluate our method at both pixel information and semantic information levels, respectively. The qualitative and quantitative experiments verify the superiority of SDCFusion in terms of visual effects and metrics. The object detection and semantic segmentation experiments demonstrate that SDCFusion achieves superior performance in high-level vision tasks. The source code is available at https://github.com/XiaoW-Liu/SDCFusion.

Improved U-net network asphalt pavement crack detection method.

Road crack detection is one of the important parts of road safety detection. Aiming at the problems such as weak segmentation effect of basic U-Net on pavement crack, insufficient precision of crack contour segmentation, difficult to identify narrow crack and low segmentation accuracy, this paper proposes an improved U-net network pavement crack segmentation method. VGG16 and Up_Conv (Upsampling Convolution) modules are introduced as backbone network and feature enhancement network respectively, and the more abstract features in the image are extracted by using the Block depth separable convolution blocks, and the multi-scale features are captured and enhanced by higher level semantic information to improve the recognition accuracy of narrow cracks in the road surface. The improved network embedded the Ca(Channel Attention) attention mechanism in U-net's jump connection to enhance the crack portion to suppress background noise. At the same time, DG_Conv(Depthwise GSConv Convolution) module and UnetUp(Unet Upsampling) module are added in the decoding part to extract richer features through more convolutional layers in the network, so that the model pays more attention to the detailed part of the crack, so the segmentation accuracy can be improved. In order to verify the model's ability to detect cracks in complex backgrounds, experiments were carried out on CFD and Deepcrack datasets. The experimental results show that compared with the traditional U-net network F1-score and mIoU have increased by 13.6% and 9.9% respectively. Superior to advanced models such as U-net, Segnet and Linknet in accuracy and generalization ability, the improved model provides a new method for asphalt pavement crack detection. The model is more conducive to practical application and ground deployment, and can be applied in road maintenance projects.

The Geometry of Low- and High-Level Perceptual Spaces.

Low-level features are typically continuous (e.g., the gamut between two colors), but semantic information is often categorical (there is no corresponding gradient between dog and turtle) and hierarchical (animals live in land, water, or air). To determine the impact of these differences on cognitive representations, we characterized the geometry of perceptual spaces of five domains: a domain dominated by semantic information (animal names presented as words), a domain dominated by low-level features (colored textures), and three intermediate domains (animal images, lightly texturized animal images that were easy to recognize, and heavily texturized animal images that were difficult to recognize). Each domain had 37 stimuli derived from the same animal names. From 13 participants (9F), we gathered similarity judgments in each domain via an efficient psychophysical ranking paradigm. We then built geometric models of each domain for each participant, in which distances between stimuli accounted for participants' similarity judgments and intrinsic uncertainty. Remarkably, the five domains had similar global properties: each required 5-7 dimensions, and a modest amount of spherical curvature provided the best fit. However, the arrangement of the stimuli within these embeddings depended on the level of semantic information: dendrograms derived from semantic domains (word, image, and lightly texturized images) were more "tree-like" than those from feature-dominated domains (heavily texturized images and textures). Thus, the perceptual spaces of domains along this feature-dominated to semantic-dominated gradient shift to a tree-like organization when semantic information dominates, while retaining a similar global geometry.

Object and attribute recognition for product image with self-supervised learning

Accurate class and attribute recognition is the critical technique to convert the unstructured product image data into structured knowledge base, which provides strong support for product design in the future. However, objects of different classes sharing similar attribute may vary a lot in visual appearances, making it challenging to accurately recognize the objects and their attributes. Different from the traditional multi-label image recognition, the attribute of the object, as high-level semantic information, is not corresponding to certain regions of the object and requires to learn more fine-grained features to represent the latent high level semantic information for the attribute across different object categories. Therefore, a self-supervised method called Deconstruction and Reconstruction Learning Network (DRLN) is proposed to solve the above problems in this paper. The DRLN tries to learn more fine-grained and local feature of the input product image by a self-supervised task, which deconstructs the input product images by randomly shuffling their local regions and further reconstructs the features of corresponding deconstructed images. Besides, the proposed model is optimized in an end-to-end manner by learning from multiple tasks, i.e., the multi-label classification task, the adversarial discrimination task, and the location alignment task. Experimental results demonstrate that the proposed method outperforms the state-of-the-arts for multi-label learning problems on both our product image dataset and another public available attribute recognition dataset. To facilitate future research in this field, all the datasets and codes are directly available online.11https://github.com/Yong-DAI/DRLN.

Research on named entity recognition of adverse drug reactions based on NLP and deep learning.

Introduction: Adverse drug reactions (ADR) are directly related to public health and become the focus of public and media attention. At present, a large number of ADR events have been reported on the Internet, but the mining and utilization of such information resources is insufficient. Named entity recognition (NER) is the basic work of many natural language processing (NLP) tasks, which aims to identify entities with specific meanings from natural language texts. Methods: In order to identify entities from ADR event data resources more effectively, so as to provide valuable health knowledge for people, this paper introduces ALBERT in the input presentation layer on the basis of the classic BiLSTM-CRF model, and proposes a method of ADR named entity recognition based on the ALBERT-BiLSTM-CRF model. The textual information about ADR on the website "Chinese medical information query platform" (https://www.dayi.org.cn) was collected by the crawler and used as research data, and the BIO method was used to label three types of entities: drug name (DRN), drug component (COM), and adverse drug reactions (ADR) to build a corpus. Then, the words were mapped to the word vector by using the ALBERT module to obtain the character level semantic information, the context coding was performed by the BiLSTM module, and the label decoding was using the CRF module to predict the real label. Results: Based on the constructed corpus, experimental comparisons were made with two classical models, namely, BiLSTM-CRF and BERT-BiLSTM-CRF. The experimental results show that the F 1 of our method is 91.19% on the whole, which is 1.5% and 1.37% higher than the other two models respectively, and the performance of recognition of three types of entities is significantly improved, which proves the superiority of this method. Discussion: The method proposed can be used effectively in NER from ADR information on the Internet, which provides a basis for the extraction of drug-related entity relationships and the construction of knowledge graph, thus playing a role in practical health systems such as intelligent diagnosis, risk reasoning and automatic question answering.

A semantic segmentation-based underwater acoustic image transmission framework for cooperative SLAM

With the development of underwater sonar detection technology, simultaneous localization and mapping (SLAM) approach has attracted much attention in underwater navigation field in recent years. But the weak detection ability of a single vehicle limits the SLAM performance in wide areas. Thereby, cooperative SLAM using multiple vehicles has become an important research direction. The key factor of cooperative SLAM is timely and efficient sonar image transmission among underwater vehicles. However, the limited bandwidth of underwater acoustic channels contradicts a large amount of sonar image data. It is essential to compress the images before transmission. Recently, deep neural networks have great value in image compression by virtue of the powerful learning ability of neural networks, but the existing sonar image compression methods based on neural network usually focus on the pixel-level information without the semantic-level information. In this paper, we propose a novel underwater acoustic transmission scheme called UAT-SSIC that includes semantic segmentation-based sonar image compression (SSIC) framework and the joint source-channel codec, to improve the accuracy of the semantic information of the reconstructed sonar image at the receiver. The SSIC framework consists of Auto-Encoder structure-based sonar image compression network, which is measured by a semantic segmentation network's residual. Considering that sonar images have the characteristics of blurred target edges, the semantic segmentation network used a special dilated convolution neural network (DiCNN) to enhance segmentation accuracy by expanding the range of receptive fields. The joint source-channel codec with unequal error protection is proposed that adjusts the power level of the transmitted data, which deal with sonar image transmission error caused by the serious underwater acoustic channel. Experiment results demonstrate that our method preserves more semantic information, with advantages over existing methods at the same compression ratio. It also improves the error tolerance and packet loss resistance of transmission.

A Multi-Granularity Heterogeneous Graph for Extractive Text Summarization

Extractive text summarization selects the most important sentences from a document, preserves their original meaning, and produces an objective and fact-based summary. It is faster and less computationally intensive than abstract summarization techniques. Learning cross-sentence relationships is crucial for extractive text summarization. However, most of the language models currently in use process text data sequentially, which makes it difficult to capture such inter-sentence relations, especially in long documents. This paper proposes an extractive summarization model based on the graph neural network (GNN) to address this problem. The model effectively represents cross-sentence relationships using a graph-structured document representation. In addition to sentence nodes, we introduce two nodes with different granularity in the graph structure, words and topics, which bring different levels of semantic information. The node representations are updated by the graph attention network (GAT). The final summary is obtained using the binary classification of the sentence nodes. Our text summarization method was demonstrated to be highly effective, as supported by the results of our experiments on the CNN/DM and NYT datasets. To be specific, our approach outperformed baseline models of the same type in terms of ROUGE scores on both datasets, indicating the potential of our proposed model for enhancing text summarization tasks.

Context-aware and co-attention network based image captioning model

ABSTRACT Previous captioning methods only rely on semantic-level information considering the similarity of features between image regions in visual space and ignoring the linguistic context incorporated in the decoder for caption generation. In this paper, a transformer-based co-attention network is proposed which uses linguistic information to capture the pairwise visual relationships among objects and significant visual features. We infer the entity words from the visual content of objects, during the caption generation process. Also, we infer the interactive words by focusing on the relationship between entity words, based on the relational context between words generated in the course of caption decoding. We use linguistic contextual information as a guiding force to discover the relationships between objects efficiently. Further, we capture both intra-modal and inter-modal interactions using the multilevel co-attention network. Our model attains 44.1/33.6 BLEU@4, 30.8/25.1 METEOR, 61.9/55.1 ROUGE, 132.1/69.8 CIDEr, and 24.1/17.8 SPICE scores on MSCOCO and Flickr30k datasets, respectively.

Multi-scale Superpixel based Hierarchical Attention model for brain CT classification

Brain CT image classification is critical for assisting brain disease diagnosis. The brain CT images contain much noisy information, and the lesions are unstable in shape and location, making the classification task more difficult when using conventional CNN models. In this paper, we propose a novel Multi-scale Superpixel based Hierarchical Attention (MSHA) model for brain CT classification by introducing the multi-scale superpixels to a hierarchical fusion structure to remove noise and help the model focus on the lesion areas. MSHA contains three modules: (1) a Semantic-level Information Extractor that extracts appearance and geometry information based on the superpixel of the image, (2) a Mixed Multi-head Attention module that obtains the mixed attention features from the semantic-level information, and (3) a Hierarchical Fusion Structure that fuses the multi-scale attention features from coarse to fine. Experiments on the brain CT dataset demonstrate the effectiveness of the proposed model.

Shape and Context-Based Recognition of Standard Mechanical Parts in CAD Models

The work here presented is part of a wider research project aimed at extracting and using in industrial applications high level semantic information from 3D product models that are described by means of their boundary representation (B-rep). The specific focus of the paper is the recognition among the components of the CAD model of an assembly of those belonging to some categories of standard parts largely employed in mechanical industry. The knowledge of these components is crucial to understand the structure of mechanical products as they have specific meaning and function. Standard parts follow international standard in shape and dimensions, and also typical mounting schemes exist concerning their use in the product assembly. These distinctive features have been exploited as a starting point to develop a multi-step recognition algorithm. It includes a shape-based and a context-based analysis both relying on the geometric and topological analysis of a CAD model. As already anticipated by Voelcker in his visionary ability to anticipate open challenges, the shape of an object alone is not enough to understand its function. Therefore, context assessment becomes crucial to validate the recognition given by the shape-based step. It allows to uniquely recognize components in mechanical CAD models, by confirming correct results, refusing the false positives, as well as choosing the correct one when the assignment is multiple. • An approach is shown for semantically enriching CAD models. • A two step process for standard parts recognition in CAD models is developed. • The recognition is rule-based and exploits engineering knowledge and shape analysis.

CapMatch: Semi-Supervised Contrastive Transformer Capsule With Feature-Based Knowledge Distillation for Human Activity Recognition.

This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data", weak" and "timecut", to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak-and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves F1 values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains F1 values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and k -nearest neighbor (KNN).

Pharmaceutical Foreign Particle Detection: An Efficient Method Based on Adaptive Convolution and Multiscale Attention

The detection of pharmaceutical product quality is indispensable and crucial in drug manufacture. In particular, the detection of liquid pharmaceutical products is generally performed manually and offline. However, this type of method is inefficient and possesses low precision due to limited personal sensing and eyesight capabilities. Since recent advanced computational intelligence (CI) algorithms have achieved great successes in computer vision, CI (e.g., deep neural networks (DNNs)) is expected to provide efficient and powerful tools that will substantially improve the decision-making of autonomous systems. Moreover, with the current automatic detection methods, it is still a challenge to detect tiny particles efficiently in liquid products. This article proposes an end-to-end deep learning (DL) method with adaptive convolution and multiscale attention to locate and classify foreign particles. First, we present a pixel-adaptive feature extraction (PAFE) method for extracting fine-grained features and reducing the intraclass disparities between particles. Following that, a multiscale attention-based feature fusion (MAFF) method, which effectively fuses the pixel-level and semantic-level information of particles, is proposed. Finally, we use a feature-selective anchor-free detection (FSAD) method to quickly detect foreign particles in liquid pharmaceuticals. To confirm the above initiatives, we validate the proposed method on a liquid pharmaceutical dataset, achieving a missed detection rate of 3.6 percent. The speed of our method is an order of magnitude faster than that of other methods, possibly reaching 15 frames per second (FPS). In addition, we test the model’s transferability on a wine dataset.

HFENet: Hierarchical Feature Extraction Network for Accurate Landcover Classification

Landcover classification is an important application in remote sensing, but it is always a challenge to distinguish different features with similar characteristics or large-scale differences. Some deep learning networks, such as UperNet, PSPNet, and DANet, use pyramid pooling and attention mechanisms to improve their abilities in multi-scale features extraction. However, due to the neglect of low-level features contained in the underlying network and the information differences between feature maps, it is difficult to identify small-scale objects. Thus, we propose a novel image segmentation network, named HFENet, for mining multi-level semantic information. Like the UperNet, HFENet adopts a top-down horizontal connection architecture while includes two improved modules, the HFE and the MFF. According to the characteristics of different levels of semantic information, HFE module reconstructs the feature extraction part by introducing an attention mechanism and pyramid pooling module to fully mine semantic information. With the help of a channel attention mechanism, MFF module up-samples and re-weights the feature maps to fuse them and enhance the expression ability of multi-scale features. Ablation studies and comparative experiments between HFENet and seven state-of-the-art models (U-Net, DeepLabv3+, PSPNet, FCN, UperNet, DANet and SegNet) are conducted with a self-labeled GF-2 remote sensing image dataset (MZData) and two open datasets landcover.ai and WHU building dataset. The results show that HFENet on three datasets with six evaluation metrics (mIoU, FWIoU, PA, mP, mRecall and mF1) are better than the other models and the mIoU is improved 7.41–10.60% on MZData, 1.17–11.57% on WHU building dataset and 0.93–4.31% on landcover.ai. HFENet can perform better in the task of refining the semantic segmentation of remote sensing images.

Design rationale representation approach based on intent-driven ontology and MBD model in design change

Aiming at the problem that the design rationale (DR) in design change has a wide range of sources and various forms, which makes designers difficult to reuse them, this paper proposes a DR representation approach for design change. Firstly, the overall three-layer structure of the approach is introduced; Use OWL to express the concepts related to the designer and product model in design change, and use SWRL and SQWRL to mine and extract useful DR information; A DR representation tool is developed based on CAD software to export the low semantic level information to the ontology as well as import the high semantic level information back to the Model Based Definition (MBD) model. The approach combines the expressing ability of ontology and visuality of 3D annotation to capture, represent and reuse DR information, which promotes the design change under the MBD environment. Finally, a stamping part is taken as an example, and designers can refer to the experience of the predecessors related to the changed parameter through using the tool.

Cross‐domain opinion classification via aspect analysis and attention sharing mechanism

AbstractThe purpose of cross‐domain opinion classification is to leverage useful information acquired from the source domain to train a classifier for opinion classification in the target domain, which has a huge amount of unlabeled data. An opinion classifier trained on a specific domain usually acts poorly, when directly employed to another domain. Annotating the data for all the domains is a laborious and costly process. The majority of available approaches are centered on identifying invariant features among domains. Unluckily, they are unable to properly capture the context within the sentences and better utilization of unlabeled data. To properly address this issue, we propose an aspect‐based attention model for cross‐domain opinion classification. By incorporating knowledge of aspects and sentences, the proposed model provides a transfer mechanism for better‐transferring opinions among domains. We introduce two learning networks, first learning network aims to recognize the shared features between domains, while the purpose of the second learning network is to extract the information from the aspects by utilizing shared words as a bridge. We benefit from BERT and bidirectional gated recurrent unit to get a deep understanding and deep level semantic information of the text. Further, the joint attention learning mechanism is performed for these two learning modules so that the aspects and sentences can impact the resulting opinion expression. In addition, we introduce a gradient reversal layer to obtain invariance features. The comprehensive experiments are performed on Amazon multidomain product datasets and show the effectiveness and significance of the proposed model over state‐of‐the‐art techniques.

Analysis and Prediction of Microblog Hot Topics Propagation Based on Deep Learning

Sina Microblog, China’s most popular social media platform, has a massive amount of data and users. The prediction of microblog forwarding has become a hot research topic in the current academic circle. The majority of current microblog forwarding research relies on traditional models that only use certain data properties or statistical features akin to term frequency-inverse document frequency (TF-IDF) for training but fail to extract microblog semantic-level information. In light of the popularity of evolutionary analysis and forecasting your own research only on particular moment popularity or predict popularity, finally, we only analyzed the influence of different factors on the final popularity, without considering the various factors in the popularity of the role played by different evolutionary stages and the popularity of evolutionary problems such as insufficient understanding. This article proposes a three-dimensional feature model. Average, trend, and cycle were used to fit and predict the prevalence, as well as the prevalence prediction method based on deep learning. We analyzed the hot topic evolvement popularity and defined a 3D character model. Regarding average, trend, and cycle and based on 3D feature model to create time series model to fit hot subject popularity evolution and forecast the short-term popularity of hot topic evolution, comparing Spike M and SH model, this work puts forward the model fit and greater accuracy. From the analysis and quantification of influencing factors in the three key stages of epidemic evolution, outbreak, peak, and decline, a prediction model of the epidemic based on a deep neural network was proposed, the different effects of each influencing factor in the evolution process of the epidemic were analyzed in detail, and the active period of the epidemic was predicted. In comparison with the Spike M and SVR models, the proposed technique has a greater effect and performance in terms of predictability and timeliness.