Computer and Modernization

Incremental News Recommendation Method Based on Self-supervised Learning and Data Replay

LIN Wei

2023, 0(12): 1-6. doi:10.3969/j.issn.1006-2475.2023.12.001

Asbtract ( 102 )

PDF (1490KB) ( 75 )

References | Related Articles | Metrics

Abstract： Personalized news recommendation technology is important to alleviate information overload and improve user experience. News recommendation models are usually iteratively trained based on fixed data sets. However， in real scenarios， news recommendation models need to constantly learn to adapt to new users and news. Therefore， incremental learning has been proposed to help models perform incremental updates. The main challenge of the incremental learning of news recommendation models is the catastrophic forgetting problem， where a trained model forgets the user preferences it has previously learned. In view of this， this paper proposes SSL-DR， an incremental learning method of news recommendation models based on self-supervised learning and data replay. SSL-DR firstly adds the self-supervised learning task to the news recommendation task to obtain the user's stable preference， which effectively reduces the problem of catastrophic forgetting. To consolidate the learned knowledge， SSL-DR further implements a sampling strategy based on the user's click probability scores for candidate news to achieve data replay and transfer the learned knowledge through a knowledge distillation strategy. The experimental results show that， our method can effectively improve the overall recommendation performance of the news recommendation model in the process of incremental training， and significantly alleviate the problem of catastrophic forgetting.

Few-shot Object Detection via Learnable Memory Feature Pyramid Network

XIA Qian-han, HE Sheng-huang, WU Yuan-qing, ZHAO Le-le

2023, 0(12): 7-13. doi:10.3969/j.issn.1006-2475.2023.12.002

Asbtract ( 51 )

PDF (2370KB) ( 64 )

References | Related Articles | Metrics

Abstract: At present， it is difficult to obtain the data of some industry application scenarios， and the problem of few shot has become an important factor restricting the application and promotion of deep learning technology. In this paper， few shot method is adopted to improve the performance of the model in the absence of data and reduce the dependence of the deep learning model on data， and few-shot object detection via learnable memory feature pyramid network is proposed to retain cleaner multi-scale feature information for classifier prediction. With the help of the adaptive feature fusion module， the network can choose the emphasis ratio among the features of different levels to maximize the retention of discriminant feature information of different scales. At the same time， we also add a retrospective feature alignment module to alleviate the feature confusion effect introduced by stacking feature layers. The experimental results show that the model performance can be effectively improved by overcoming the dependence on data， and the improved model can surpass other existing models of the same type in the COCO dataset and VOC dataset. In particular， when the prior parameter k is set to 5 in VOC dataset， nAP50 increases by 4.8 to 44.7； when the prior parameter k is set to 30 in COCO dataset， nAP50 increases by 4.0 to 29.4.

Sliding Mode Control System of Manipulator Based on Improved#br# Variable Structure Reaching Law#br#

SONG Tao-tao, LI Yan-ping, LI Hong-gang, HAN Chun-xue

2023, 0(12): 14-18. doi:10.3969/j.issn.1006-2475.2023.12.003

Asbtract ( 41 )

PDF (3989KB) ( 52 )

References | Related Articles | Metrics

Abstract： Aiming at the problems of slow convergence speed and system chattering in the application field of sliding mode control in the manipulator， an improved variable structure reaching law is designed to improve the dynamic characteristics of the manipulator. The convergence speed and chattering suppression of the system are optimized by using the grouping characteristics and the inverse hyperbolic sine function. The mathematical model of the two degree of freedom manipulator is established by using the second Lagrange equation， and the RBF neural network is used to approximate the system model for the friction and other unmeasurable interference problems in the system. The stability of the system tracking is proved by Lyapunov function method. Finally， the feasibility and stability of the improved variable structure reaching law algorithm are verified by the experimental comparison with PID， constant rate reaching law and fast power reaching law in Simulink.

EEG Recognition of Motor Imagination Based on Efficiency Channel Attention Module

ZHOU Cheng-cheng, ZENG Qing-jun, YANG Kang, HU Jia-ming, HAN Chun-wei

2023, 0(12): 19-23. doi:10.3969/j.issn.1006-2475.2023.12.004

Asbtract ( 57 )

PDF (1293KB) ( 69 )

References | Related Articles | Metrics

Abstract： The brain-computer interface technology based on motor imagination is helpful to the rehabilitation of patients with hand movement disorders， so it is widely used in the field of rehabilitation medicine. Aiming at the problem of poor classification of motor imagination-electroencephalogram （MI-EEG） due to its low signal-to-noise ratio in current motor imagination-electroencephalogram， in view of the ability of the attention module to focus on important features related to motor imagination classification tasks and ignore unimportant features， we propose a convolutional neural network based on the efficient channel attention （ECA） module for feature extraction and classification of left and right-handed MI-EEG. In order to facilitate the recognition of EEG signals by convolutional neural network （CNN）， this paper uses wavelet transform to convert the timing signals of C3 and C4 channels into two-dimensional time-frequency graphs， then designs a CNN structure and parameters based on ECA. Finally， the proposed method is tested on EEG data set. The experimental results show that compared with CNN and the CNN method based on fusion convolution attention， the CNN method based on ECA can effectively improve the recognition accuracy of MI-EEG， indicating that the proposed method is effective in motor imagination EEG recognition.

Point Cloud Completion Algorithm Based on Multi-stage Fractal Combination

ZENG Wei-ping, CHEN Jun-hong, Muhammad ASIM, LIU Wen-yin, YANG Zhen-guo

2023, 0(12): 24-29. doi:10.3969/j.issn.1006-2475.2023.12.005

Asbtract ( 49 )

PDF (1816KB) ( 50 )

References | Related Articles | Metrics

Abstract: Point cloud is a common representation of 3D objects. However， due to reasons such as sensor design and precision， the obtained point cloud usually has missing geometry and sparseness. To solve this problem， this paper proposes a point cloud completion algorithm based on multi-stage fractal combination. In the first stage， the input point cloud is sampled multiple times and features are extracted separately， then the pyramid model is used to generate a point cloud with multi-scale geometry loss， and finally， the generated point cloud is spliced with the input point cloud. In the second stage， KNN clustering and PointNet stacking network are used to extract local features， and the spliced point cloud is down-sampled as a rough prediction， and finally， the rough prediction is combined with the local input folding network to generate a refined high-quality point cloud. This algorithm is based on local to overall multi-stage completion， and the loss function can be adjusted for different stages， which effectively optimizes the completion process and achieves good completion results in the ShapeNet dataset.

Improved Harris Hawks Optimization Algorithm Based on Cluster Centroid and#br# Exponential Decline Method#br#

BAI Xiao-bo, JIANG Meng-xi, WANG Tie-shan, SHAO Jing-feng, LI Bo,

2023, 0(12): 30-35. doi:10.3969/j.issn.1006-2475.2023.12.006

Asbtract ( 28 )

PDF (1730KB) ( 36 )

References | Related Articles | Metrics

Abstract: To promote optimization performance of Harris hawks optimization algorithm， KmHHO algorithm is proposed. Firstly， all populations as a cluster， the cluster centroid is calculated with Kmeans of Matlab， mean of HHO is replaced by cluster centroid. Then， to control the segments of exploration and development， linearly decreasing escape energy of prey is replaced with exponentially decreasing escape energy of prey. Finally， searching performance of five algorithms is compared on 23 benchmark functions， the improved effect of KmHHO is verified and Wilcoxon rank sum test is utilized to analyze the difference of KmHHO with other four optimization algorithms. The experimental results show that among the 23 benchmarks， KmHHO can achieve the optimal value on 14 benchmark functions， and its overall performance is higher than GWO， HHO and AO， but it’s equivalent to DAHHO.

Deep Federated Image Classification Method Based on Bilateral Homomorphic Encryption

LIANG Tian-kai, HUANG Kang-hua, LIU Kai-hang, LAN Lan, ZENG Bi

2023, 0(12): 36-40. doi:10.3969/j.issn.1006-2475.2023.12.007

Asbtract ( 61 )

PDF (2194KB) ( 58 )

References | Related Articles | Metrics

Abstract： Concerning the privacy protection and data island problems of traditional machine learning paradigm， combined with deep learning， a deep federated image classification method based on bilateral homomorphic encryption called AFL algorithm is proposed. Firstly， AFL algorithm is a horizontal federated improvement of the VGG neural network. At the same time， a bi-directional Paillier homomorphic encryption mechanism based on the Paillier homomorphic encryption algorithm called Bi-HE mechanism is proposed， which can ensure the privacy and security of the federated system. Secondly， the AFL algorithm proposes an adaptive waiting strategy during model aggregation， which can effectively avoids the problem of low aggregation efficiency caused by communication blockage. Finally， the experiments using the CIFAR-10 data set have proved that the AFL algorithm has better generalization capabilities which can effectively solve the problems of privacy protection and data islands compared with the traditional VGG and DenseNet algorithms， and the AFL algorithm is better than the traditional federated learning model in efficiency.

A Multi-label Image Classification Model Based on Dual Feature Attention

QIU Kai-xing, FENG Guang

2023, 0(12): 41-47. doi:10.3969/j.issn.1006-2475.2013.12.008

Asbtract ( 17 )

PDF (1495KB) ( 23 )

References | Related Articles | Metrics

Abstract： A multi-label image classification model based on dual feature attention is proposed to address the current problems of insufficient extraction of feature information from multiple image regions and difficulty in constructing semantic relationships between image features and labels in multi-label image classification tasks. Firstly, the image feature attention module is constructed to correlate the attention of image information with global multi-region features to enhance image feature extraction. Secondly, a combined feature attention module is constructed to perform correlation representation of image feature information and label embedding, thus enabling cross-modal fusion between labels and image regions to obtain a better mapping relationship. The experimental results show that the model achieves better classification results in both the VOC2007 and COCO2014 multi-label image classification datasets, and its performance metrics have improved significantly compared with existing algorithms, verifying the effectiveness of the model.

Infrared Spectrum Modeling Method Based on Variable Selection of Model#br# Population Analysis#br#

DU Kang, GUO Lu-yu, XU Qi-lei, SHAN Bao-ming, ZHANG Fang-kun

2023, 0(12): 48-52. doi:10.3969/j.issn.1006-2475.2023.12.009

Asbtract ( 37 )

PDF (2663KB) ( 34 )

References | Related Articles | Metrics

Abstract： The variable selection method can realize the dimensionality reduction of high-dimensional data， reduce the complexity of the calibration model as well as improve the predictive ability and interpretability of the model， which is important for establishing an efficient and reliable prediction model. In this paper， model population analysis （MPA） is used for variable selection in the modeling process of NIR spectral calibration. A subset index reuse kernel - partial least squares （SIRK-PLS） fusion modeling approach is proposed by combining the characteristics of MPA to repeatedly extract subsets in the same space. The method essentially avoids redundant calculations in the process of cross-validation of variable selection subsets and regression coefficient solving under the MPA framework by indexing the pre-calculated covariance matrix， and improves modeling efficiency. In addition， the SIRK-PLS modeling approach allows for automatic optimal switching of modeling algorithms based on the ratio of the number of samples to the number of variables. The algorithm performance is validated with a nominal near-infrared spectral corn data set. The results show that the SIRK-PLS modeling method proposed in this paper has fast convergence speed and high accuracy， and is suitable for automatic and fast dimensionality reduction modeling of mobile infrared spectroscopy devices， which has some application prospects.

Course Recommendation Method Combining Time Correlation Degree and Course#br# Collocation Degree

LIU Yu-cheng, HE Qi, DONG Yan-hua, WANG Xiao-yu

2023, 0(12): 53-58. doi:10.3969/j.issn.1006-2475.2023.12.010

Asbtract ( 36 )

PDF (1819KB) ( 46 )

References | Related Articles | Metrics

Abstract: In view of the problems in the existing course recommendation system， such as over-reliance on users' grades of courses， failure to consider the change of users' interests over time and neglect the collocation between the courses learned by users and the recommended courses， a TIMR course recommendation model based on the degree of time correlation and course collocation is proposed. On the one hand， TIMR model uses course viewing progress instead of course rating， and applies time correlation function to calculate the similarity between courses. On the other hand， the course collocation degree function is constructed by using the co-selected frequency of the course. Then， time correlation and course collocation are combined to produce predictive grades. In order to verify the validity of TIMR model， experiments are conducted on TM data set， CN data set and MOOC data set. Experiments show that compared with the existing recommendation methods UserCF， ItemCF， LFM， PR， MPR and SMCR， TIMR significantly improves the Precision， Recall and F1_score indexes， which has obvious advantages in improving the recommendation quality.

3D-SPRNet: Segmentation Model of Gallbladder Cancer Based on Parallel Decoder and Double Attention Mechanism

ZHANG Hao-yang, YIN Zi-ming, LE Jun-yi, SHEN Da-cong, SHU Yi-jun, YANG Zi-yi,

2023, 0(12): 59-66. doi:10.3969/j.issn.1006-2475.2023.12.011

Asbtract ( 47 )

PDF (1780KB) ( 55 )

References | Related Articles | Metrics

Abstract: The segmentation of cancerous part of gallbladder CT based on deep learning could be used as a diagnostic reference for clinicians. In existing methods， two-dimensional image slices that lack spatial context information are universally adopted as input. Meanwhile， the boundary segmentation is not accurate enough because of lacking the refinement of the cancer boundary region. In order to increase the accuracy of boundary segmentation and guarantee the continuity of spatial information， a 3D-SPRNet segmentation model for gallbladder carcinoma is proposed. A parallel decoder is used to extract and decode multi-scale advanced features. Channel attention is used to help network emphasize feature extraction information. Reverse attention is used to focus on the unpredicted region and gradually refine the cancer boundary. The CT images of 304 patients with gallbladder cancer from Xinhua Hospital Affiliated to Shanghai Jiaotong University School of Medicine are selected for the experiment. The MIoU， IoU and Dice coefficients obtained are 0.85， 0.70 and 0.83， respectively， which are better than those of most mainstream segmentation networks. The effectiveness of each module has been verified by ablation experiment. The experimental results show that the network model proposed in this paper can improve the problem of rough segmentation boundaries and increase the segmentation accuracy of gallbladder carcinoma.

White Matter Hyperintensities Segmentation Based on High Gray Value#br# Attention Mechanism

ZHANG Bo-quan, MAI Hai-peng, CHEN Jia-min, Pang Jin-ju

2023, 0(12): 67-75. doi:10.3969/j.issn.1006-2475.2023.12.012

Asbtract ( 40 )

PDF (2586KB) ( 50 )

References | Related Articles | Metrics

Abstract： White matter hyperintensities， commonly seen in the image of cerebral small vessel disease （CSVD）， shed light on the clinical diagnoses of patients with cerebral small vessel disease. White matter hyperintensities segmentation， as a basic work in clinical diagnosis， often requires experienced doctors to carry it out manually， which is time-consuming and intricate. White matter hyperintensities， referring to the hyperintense shadows in T2 weighted magnetic resonance images of the brains or fluid-attenuated inversion recovery sequence images， are of higher gray values than other brain tissues. To enhance the attention to areas of white matter hyperintensities， this paper proposes a network model of a high gray value attention mechanism in light of the imaging characteristics of white matter hyperintensities. The model， based on the UNet， introduces a module of high gray value attention so that it can pay more attention to the areas of relatively high gray values in the images. It also introduces a residual mixed attention module to enhance the ability for extracting features of the net model. As a result， it significantly enhances the segmentation effect of white matter hyperintensities， with its DSC and Recall indicators reaching 0.8330 and 0.8870， respectively， which is better than existing algorithms. Moreover， ablation experiments verified the effectiveness of the high gray value attention module and the residual hybrid attention module. This paper provides a new method for the FLAIR-based segmentation of white matter hyperintensities lesion， and verifies the feasibility of combining the traditional method for image segmentation with in-depth learning technology.

Application of Neural Rendering Based Visual Synthesis in Construction Scene

ZHANG Zai-cheng, LI Jian

2023, 0(12): 76-81. doi:10.3969/j.issn.1006-2475.2023.12.013

Asbtract ( 34 )

PDF (1363KB) ( 39 )

References | Related Articles | Metrics

Abstract: When neural radiation field （NeRF） is applied to outdoor construction scenes， due to the difficulty in capturing the lighting in outdoor scenes and the large difference between the foreground and background ranges of construction scenes， blurring and artifacts will appear in novel views. Through analysis， an improved visual synthesis method is proposed. First， camera parameters are obtained from RGB images through SFM algorithm to represent outdoor construction scenes. Then， the vector generated from the pre training encoder is introduced and added to the rendering network to reduce the impact of light. Finally， the foreground and background in the image are separated for volume rendering， which improves the effect of visual synthesis. Based on the outdoor construction scene data set， compared with three methods， the results show that the proposed method improves the peak signal-to-noise ratio and structural similarity by 12.2% and 10.9% respectively compared with the best method. It appears that the novel views generated by the proposed method in outdoor construction scenes have better finesse.

Application of Improved YOLOv7 Algorithm in Detection of Capping Defects of Vials

NING Juan, ZHOU Qing-hua, ZENG Xiao-wei

2023, 0(12): 82-86. doi:10.3969/j.issn.1006-2475.2023.12.014

Asbtract ( 52 )

PDF (2347KB) ( 65 )

References | Related Articles | Metrics

Abstract: Aiming at the problem of missing inspection caused by small target defects and unclear features in the defect detection of the capping of penicillin bottles， this paper proposes a defect detection method based on the improved YOLOv7 algorithm. First， the defect images of the capping of penicillin bottles are collected in the real industrial environment， including four common defects: scratches， missing cap， concave-crack， and composite defect， and the data are enhanced to construct a data set with 3220 images of penicillin bottle capping defects. Then CBAM （convolution block attention module） and ASFF （adaptive spatial feature fusion） adaptive feature fusion modules are introduced on the basis of the original YOLOv7 in order to improve the ability of network to extract features， improve the detection accuracy of small target defects， and reduce the missed detection rate of penicillin bottle capping defects. The experimental results show that the average detection accuracy （mAP） of the improved algorithm reaches 99.3%， which is 1.9 percentage points higher than that before the improvement. The improved algorithm provides a new idea for the detection of capping defects of penicillin bottles in industry， and has good application prospect.

Hand Hygiene Action Quality Assessment Based on Multi-source Action Information

LI De-kang, TANG Jin, WANG Fu-tian, TU Zi-jian,

2023, 0(12): 87-93. doi:10.3969/j.issn.1006-2475.2023.12.015

Asbtract ( 27 )

PDF (2015KB) ( 36 )

References | Related Articles | Metrics

Abstract： The research on hand hygiene action quality assessment plays a crucial role in the intervention and improvement of hand hygiene behaviors. To address this task， this paper takes two different types of hand hygiene action information from video data and differential image data as inputs， creating a hand hygiene action quality assessment method based on multi-source action information. The algorithm consists of an action segmentation module and an evaluation module. In the action segmentation module， the feature segments associated with each step are divided by the position index. In the evaluation module， the differential image features obtained by the inter-frame difference method and ResNet50 feature extractor are introduced to combine with the past method （combining the optical flow and the I3D feature information of RGB） to capture the subtle hand motion information. The feature segments obtained by borrowing the segmentation module are processed and input to the hand hygiene information decoder based on the cross-attention mechanism， and the comprehensive features that fuse the details of hand motion are obtained. Next， these features are used to calculate the evaluation score of each step， and finally the evaluation score of each step is added to obtain the final evaluation result. The algorithm is verified by using the public data set HHA300. In the evaluation task， the evaluation index ρ and R-[ℓ]2（×100）achieves 0.86 and 0.95 respectively， which fully proves that the algorithm can accurately evaluate the motion quality of hand hygiene.

DOA Estimation of One-dimensional Hybrid Sensors Array Signals

HU Bi-wei, WU Zhi-ping, RAO Wei, LI Yuan-qing, HU Bi-wei

2023, 0(12): 94-99. doi:10.3969/j.issn.1006-2475.2023.12.016

Asbtract ( 24 )

PDF (1788KB) ( 25 )

References | Related Articles | Metrics

Abstract： The vector sensors can provide more available information， e.g.， acoustic information and polarization information， so the vector sensor arrays outperform the scalar sensor arrays in direction of arrival （DOA） estimation. However， the cost of vector sensor arrays is higher than that of scalar sensor arrays due to the complexity of constructing single vector sensors by itself. To reduce the cost of the vector sensor array， a hybrid sensor array consisting of acoustic vector sensors and scalar sensors is proposed firstly. And then by using the tensor algebra， a third-order data tensor can be obtained from the second-order statistics of the received signal from the proposed array. This tensor model corresponds to a virtual uniform linear array of acoustic vector sensors， allowing the proposed hybrid sensor array to have estimation performance beyond the same number of acoustic vector sensor arrays at a lower cost. Analyses show that a virtual uniform linear array with approximately 2M[22]-M2 acoustic vector sensors can be obtained from the proposed array with M2 acoustic vector sensors and 2M2 scalar sensors. Finally， based on this tensor model， the tensor decomposition can be used to achieve DOA estimation. Under the same array cost， the proposed method outperforms related methods in estimating DOA. Simulation results show the effectiveness of the proposed method.

Dynamic Threat Assessment of Air Swarm Targets Based on Intent Recognition

WANG Yu-hang, DONG Bao-liang, GONG Chao, SHANG Zhen-zhen, YAO Kang-ning

2023, 0(12): 100-104. doi:10.3969/j.issn.1006-2475.2023.12.017

Asbtract ( 27 )

PDF (1265KB) ( 50 )

References | Related Articles | Metrics

Abstract： In order to solve the problem of the decline of evaluation accuracy caused by the ignorance of situation elements with time by traditional threat assessment algorithms， this paper proposes a dynamic threat assessment method for air swarm targets based on intent recognition. In this method， the Long Short-Term Memory （LSTM） network is first used for intention prediction， and then the attention mechanism is used to improve the feature learning ability of the intention prediction model， and the multi-dimensional features of the input are weighted to a certain extent， so that the degree of influence of different features on the results is different. Softmax is used to classify the intention results， and then the results of intention prediction are used as important inputs for threat assessment in a cascading manner. Combined with static situation elements and dynamic situation elements at the current moment， multi-layer perceptron （MLP） is used for threat assessment. Simulation experiments show that compared with the traditional threat assessment method， the dynamic threat assessment method for air swarm targets based on intent recognition is more accurate.

Optimal Configuration of Multi-energy Mobile Power Vehicles Based on Improved GSA Algorithm

WANG Kai-xiang, YANG Jing, YANG Wen, MI Hong-ju, GAN Fei

2023, 0(12): 105-111. doi:10.3969/j.issn.1006-2475.2023.12.018

Asbtract ( 19 )

PDF (2415KB) ( 33 )

References | Related Articles | Metrics

Abstract: The traditional energy supply mode is difficult to cover the energy islands in the plateau and cold regions， and the multi-energy mobile power vehicle has become a better solution because of its flexible mobility and strong environmental adaptability. The existing multi-energy mobile power vehicles still lack application research for the unique background of plateau and cold， existing multi-energy allocation algorithms have some problems such as slow convergence speed and easy to fall into local optimality. A multi-energy configuration algorithm is proposed on the basis of the improved universal gravitational search algorithm in this work. In order to decrease the annual economic cost of multi-energy mobile power vehicles， the particle swarm optimization algorithm is introduced in this modified algorithm on the basis of the universal gravitational algorithm. Meanwhile， the individual historical optimal and global optimal position assignment values are introduced for particle swarm velocity iteration calculation， thus improving the speed and directionality of particle swarm convergence. The superiority of the algorithm in convergence speed and global search capability is verified by practical application cases in the Somewhere region of Tibet. The results show that the multi-energy configuration strategy of mobile power vehicles designed based on the proposed algorithm has better economy and can provide a design basis for the optimal configuration of multi-energy mobile power vehicles in highland alpine areas.

Fault Detection Method of Control Cluster Based on AOA-MSVM

YANG Bo, ZHUANG Yi

2023, 0(12): 112-116. doi:10.3969/j.issn.1006-2475.2023.12.019

Asbtract ( 23 )

PDF (1763KB) ( 27 )

References | Related Articles | Metrics

Abstract： The complexity and diversity of the control cluster system lead to the easy failure of the cluster system，which reduces the availability of the cluster system. In view of the problems in the existing fault detection methods，such as low detection efficiency and accuracy，and difficulty in effective automatic identification of fault types，in this paper，a control cluster fault detection method based on an improved adaptive Arithmetic Optimization Algorithm-Multi-class SVM （AOA-MSVM） is proposed to detect cluster faults in order to improve the availability of the cluster system. Firstly，the local linear embedding algorithm is used to reduce the dimensionality of the system information detected in the cluster system. Then，according to the characteristics of multiple kinds of faults in the cluster system，the method of one-to-many support vector machine is used to build a fault detection model to improve the ability of fault detection. Finally，the improved adaptive arithmetic optimization algorithm is used to obtain the optimal solution of the model parameters. A high availability control cluster system is set up for comparative experiments. The experimental results show that the proposed fault detection method has higher detection efficiency and accuracy and can effectively identify the fault type.

Design of Hybrid Workload-oriented Distributed Meteorological Data Management System

CHEN Chao, GU Qing-feng

2023, 0(12): 118-122. doi:10.3969/j.issn.1006-2475.2023.12.020

Asbtract ( 24 )

PDF (2516KB) ( 37 )

References | Related Articles | Metrics

Abstract: Meteorological data has the characteristics of large data scale and diverse data types. It has high requirements for access performance， so a high-performance distributed data management system is highly required. However， the meteorology is not a pure OLTP or a pure OLAP application， but a combination of the two， i.e.， HTAP with both a large number of data updating and highly concurrent data queries. Although distributed data management systems are rapidly evolving nowadays， their current support for HTAP is not very good. Therefore， in this paper， we design and implement a new hybrid workload-oriented distributed meteorological data management system for meteorological data， especially for the large-scale and most frequently used meteorological model data. The heterogeneity of different types of storage models， i.e.， grid-based and priority-based storage models， in this system can satisfy all the requirements of different types of complex queries efficiently for higher overall performance. This system can promote the concurrent query performance by 3.13 times under similar writing performance.

Table of Content