In the market of agricultural products, the price of agricultural products is affected by production cost, market supply and other factors. In order to obtain the market information of agricultural products, the price...In the market of agricultural products, the price of agricultural products is affected by production cost, market supply and other factors. In order to obtain the market information of agricultural products, the price fluctuation can be analyzed and predicted. A distributed big data software platform based on Hadoop, Hive and Spark is proposed to analyze and forecast agricultural price data. Firstly, Hadoop, Hive and Spark big data frameworks were built to store the data information of agricultural products crawled into MYSQL. Secondly, the information of agricultural products crawled from MYSQL was exported to a text file, uploaded to HDFS, and mapped to spark SQL database. The data was cleaned and improved by Holt-Winters (three times exponential smoothing method) model to predict the price of agricultural products in the future. The data cleaned by spark SQL was imported and predicted by improved Holt-Winters into MYSQL database. The technologies of pringMVC, Ajax and Echarts were used to visualize the data.展开更多
Aiming at geological disaster monitoring and prevention work, a real-time monitoring and early warning system is proposed for low-power consumption of landslides to meet the needs of landslide monitoring in remote mou...Aiming at geological disaster monitoring and prevention work, a real-time monitoring and early warning system is proposed for low-power consumption of landslides to meet the needs of landslide monitoring in remote mountainous areas. The inclination angle of the mountain body was detected by a mechanical inclination sensor, and a plurality of inclination sensors were placed on each landslide body to form an array distribution. The landslide body was stereoscopically monitored. Each mesh node had a different node address, and different inclination thresholds were set in advance. When the sensor detection value reached the threshold, an alarm message was sent to the system main control end, and the main control end generated an audible and visual alarm, and an alarm message was sent at the same time. Compared with the current landslide warning system on the market, the system achieves expected results and its power consumption is extremely low. The sensor terminal of the mountain monitoring is powered by dry battery and can work for 6 years in the field without external power supply. It avoids the damages made by weather, livestock and human, and has broad application prospects.展开更多
The high-frequency(HF) communication is one of essential communication methods for military and emergency application. However, the selection of communication frequency channel is always a difficult problem as the cro...The high-frequency(HF) communication is one of essential communication methods for military and emergency application. However, the selection of communication frequency channel is always a difficult problem as the crowded spectrum, the time-varying channels, and the malicious intelligent jamming. The existing frequency hopping, automatic link establishment and some new anti-jamming technologies can not completely solve the above problems. In this article, we adopt deep reinforcement learning to solve this intractable challenge. First, the combination of the spectrum state and the channel gain state is defined as the complex environmental state, and the Markov characteristic of defined state is analyzed and proved. Then, considering that the spectrum state and channel gain state are heterogeneous information, a new deep Q network(DQN) framework is designed, which contains multiple sub-networks to process different kinds of information. Finally, aiming to improve the learning speed and efficiency, the optimization targets of corresponding sub-networks are reasonably designed, and a heterogeneous information fusion deep reinforcement learning(HIF-DRL) algorithm is designed for the specific frequency selection. Simulation results show that the proposed algorithm performs well in channel prediction, jamming avoidance and frequency channel selection.展开更多
With the popularity of online learning and due to the significant influence of emotion on the learning effect,more and more researches focus on emotion recognition in online learning.Most of the current research uses ...With the popularity of online learning and due to the significant influence of emotion on the learning effect,more and more researches focus on emotion recognition in online learning.Most of the current research uses the comments of the learning platform or the learner’s expression for emotion recognition.The research data on other modalities are scarce.Most of the studies also ignore the impact of instructional videos on learners and the guidance of knowledge on data.Because of the need for other modal research data,we construct a synchronous multimodal data set for analyzing learners’emotional states in online learning scenarios.The data set recorded the eye movement data and photoplethysmography(PPG)signals of 68 subjects and the instructional video they watched.For the problem of ignoring the instructional videos on learners and ignoring the knowledge,a multimodal emotion recognition method in video learning based on knowledge enhancement is proposed.This method uses the knowledge-based features extracted from instructional videos,such as brightness,hue,saturation,the videos’clickthrough rate,and emotion generation time,to guide the emotion recognition process of physiological signals.This method uses Convolutional Neural Networks(CNN)and Long Short-Term Memory(LSTM)networks to extract deeper emotional representation and spatiotemporal information from shallow features.The model uses multi-head attention(MHA)mechanism to obtain critical information in the extracted deep features.Then,Temporal Convolutional Network(TCN)is used to learn the information in the deep features and knowledge-based features.Knowledge-based features are used to supplement and enhance the deep features of physiological signals.Finally,the fully connected layer is used for emotion recognition,and the recognition accuracy reaches 97.51%.Compared with two recent researches,the accuracy improved by 8.57%and 2.11%,respectively.On the four public data sets,our proposed method also achieves better results compared with the two recent researches.The experiment results show that the proposed multimodal emotion recognition method based on knowledge enhancement has good performance and robustness.展开更多
Predicting the usage of container cloud resources has always been an important and challenging problem in improving the performance of cloud resource clusters.We proposed an integrated prediction method of stacking co...Predicting the usage of container cloud resources has always been an important and challenging problem in improving the performance of cloud resource clusters.We proposed an integrated prediction method of stacking container cloud resources based on variational modal decomposition(VMD)-Permutation entropy(PE)and long short-term memory(LSTM)neural network to solve the prediction difficulties caused by the non-stationarity and volatility of resource data.The variational modal decomposition algorithm decomposes the time series data of cloud resources to obtain intrinsic mode function and residual components,which solves the signal decomposition algorithm’s end-effect and modal confusion problems.The permutation entropy is used to evaluate the complexity of the intrinsic mode function,and the reconstruction based on similar entropy and low complexity is used to reduce the difficulty of modeling.Finally,we use the LSTM and stacking fusion models to predict and superimpose;the stacking integration model integrates Gradient boosting regression(GBR),Kernel ridge regression(KRR),and Elastic net regression(ENet)as primary learners,and the secondary learner adopts the kernel ridge regression method with solid generalization ability.The Amazon public data set experiment shows that compared with Holt-winters,LSTM,and Neuralprophet models,we can see that the optimization range of multiple evaluation indicators is 0.338∼1.913,0.057∼0.940,0.000∼0.017 and 1.038∼8.481 in root means square error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE)and variance(VAR),showing its stability and better prediction accuracy.展开更多
The recommendation system(RS)on the strength of Graph Neural Networks(GNN)perceives a user-item interaction graph after collecting all items the user has interacted with.Afterward the RS performs neighborhood aggregat...The recommendation system(RS)on the strength of Graph Neural Networks(GNN)perceives a user-item interaction graph after collecting all items the user has interacted with.Afterward the RS performs neighborhood aggregation on the graph to generate long-term preference representations for the user in quick succession.However,user preferences are dynamic.With the passage of time and some trend guidance,users may generate some short-term preferences,which are more likely to lead to user-item interactions.A GNN recommendation based on long-and short-term preference(LSGNN)is proposed to address the above problems.LSGNN consists of four modules,using a GNN combined with the attention mechanism to extract long-term preference features,using Bidirectional Encoder Representation from Transformers(BERT)and the attention mechanism combined with Bi-Directional Gated Recurrent Unit(Bi-GRU)to extract short-term preference features,using Convolutional Neural Network(CNN)combined with the attention mechanism to add title and description representations of items,finally inner-producing long-term and short-term preference features as well as features of items to achieve recommendations.In experiments conducted on five publicly available datasets from Amazon,LSGNN is superior to state-of-the-art personalized recommendation techniques.展开更多
A five-axis camera stabilizer based on quaternion unscented Kalman filter algorithm is designed. It combined the unscented Kalman filter algorithm with the quaternion attitude solution and was solved by attitude senso...A five-axis camera stabilizer based on quaternion unscented Kalman filter algorithm is designed. It combined the unscented Kalman filter algorithm with the quaternion attitude solution and was solved by attitude sensor. By attitude algorithm, the motor in three directions of pitch, heading and roll in the stabilizer was accurately adjusted to control the movement of the three electronic arms. In order to improve the three-axis hand-held camera stabilizer’s performance, and to solve the jitter problem of up-and-down movement not being eliminated, two mechanical anti-shake arms were loaded under the stabilizer to balance the camera’s picture in pitch, roll, heading, and above and below five directions. Movement can maintain a stable effect. The simulation results show that the algorithm can effectively suppress the attitude angle divergence and improve the attitude calculation accuracy.展开更多
Due to the development of E-Commerce, collaboration filtering (CF) recommendation algorithm becomes popular in recent years. It has some limitations such as cold start, data sparseness and low operation efficiency. In...Due to the development of E-Commerce, collaboration filtering (CF) recommendation algorithm becomes popular in recent years. It has some limitations such as cold start, data sparseness and low operation efficiency. In this paper, a CF recommendation algorithm is propose based on the latent factor model and improved spectral clustering (CFRALFMISC) to improve the forecasting precision. The latent factor model was firstly adopted to predict the missing score. Then, the cluster validity index was used to determine the number of clusters. Finally, the spectral clustering was improved by using the FCM algorithm to replace the K-means in the spectral clustering. The simulation results show that CFRALFMISC can effectively improve the recommendation precision compared with other algorithms.展开更多
Fuzzy C-Means(FCM)is an effective and widely used clustering algorithm,but there are still some problems.considering the number of clusters must be determined manually,the local optimal solutions is easily influenced ...Fuzzy C-Means(FCM)is an effective and widely used clustering algorithm,but there are still some problems.considering the number of clusters must be determined manually,the local optimal solutions is easily influenced by the random selection of initial cluster centers,and the performance of Euclid distance in complex high-dimensional data is poor.To solve the above problems,the improved FCM clustering algorithm based on density Canopy and Manifold learning(DM-FCM)is proposed.First,a density Canopy algorithm based on improved local density is proposed to automatically deter-mine the number of clusters and initial cluster centers,which improves the self-adaptability and stability of the algorithm.Then,considering that high-dimensional data often present a nonlinear structure,the manifold learning method is applied to construct a manifold spatial structure,which preserves the global geometric properties of complex high-dimensional data and improves the clustering effect of the algorithm on complex high-dimensional datasets.Fowlkes-Mallows Index(FMI),the weighted average of homogeneity and completeness(V-measure),Adjusted Mutual Information(AMI),and Adjusted Rand Index(ARI)are used as performance measures of clustering algorithms.The experimental results show that the manifold learning method is the superior distance measure,and the algorithm improves the clustering accuracy and performs superiorly in the clustering of low-dimensional and complex high-dimensional data.展开更多
Container virtual technology aims to provide program independence and resource sharing.The container enables flexible cloud service.Compared with traditional virtualization,traditional virtual machines have difficulty...Container virtual technology aims to provide program independence and resource sharing.The container enables flexible cloud service.Compared with traditional virtualization,traditional virtual machines have difficulty in resource and expense requirements.The container technology has the advantages of smaller size,faster migration,lower resource overhead,and higher utilization.Within container-based cloud environment,services can adopt multi-target nodes.This paper reports research results to improve the traditional trust model with consideration of cooperation effects.Cooperation trust means that in a container-based cloud environment,services can be divided into multiple containers for different container nodes.When multiple target nodes work for one service at the same time,these nodes are in a cooperation state.When multi-target nodes cooperate to complete the service,the target nodes evaluate each other.The calculation of cooperation trust evaluation is used to update the degree of comprehensive trust.Experimental simulation results show that the cooperation trust evaluation can help solving the trust problem in the container-based cloud environment and can improve the success rate of following cooperation.展开更多
There is a large amount of information in the network data that we canexploit. It is difficult for classical community detection algorithms to handle network data with sparse topology. Representation learning of netw...There is a large amount of information in the network data that we canexploit. It is difficult for classical community detection algorithms to handle network data with sparse topology. Representation learning of network data is usually paired with clustering algorithms to solve the community detection problem.Meanwhile, there is always an unpredictable distribution of class clusters outputby graph representation learning. Therefore, we propose an improved densitypeak clustering algorithm (ILDPC) for the community detection problem, whichimproves the local density mechanism in the original algorithm and can betteraccommodate class clusters of different shapes. And we study the communitydetection in network data. The algorithm is paired with the benchmark modelGraph sample and aggregate (GraphSAGE) to show the adaptability of ILDPCfor community detection. The plotted decision diagram shows that the ILDPCalgorithm is more discriminative in selecting density peak points compared tothe original algorithm. Finally, the performance of K-means and other clusteringalgorithms on this benchmark model is compared, and the algorithm is proved tobe more suitable for community detection in sparse networks with the benchmarkmodel on the evaluation criterion F1-score. The sensitivity of the parameters ofthe ILDPC algorithm to the low-dimensional vector set output by the benchmarkmodel GraphSAGE is also analyzed.展开更多
Pedestrian detection has a wide range of applications in daily life, and many fields require pedestrians to conduct detection with high precision and speed, which is an urgent problem to be solved. The traditional ped...Pedestrian detection has a wide range of applications in daily life, and many fields require pedestrians to conduct detection with high precision and speed, which is an urgent problem to be solved. The traditional pedestrian detection method improves the detection performance by improving the classification algorithm and extracting more effective features. In this paper, a pedestrian detection method is proposed based on single shot multibox detector (SSD) model, which replaces the basic network part of SSD model with inception network structure with smaller parameters, faster running speed and stronger nonlinear expression ability. A high-performance network model for pedestrian detection was based on improved SSD. The experimental results show that the proposed method is faster than the original model, and the average precision of pedestrian recognition and location is 89.6%, which is 2.6% higher than the original model.展开更多
As the technological breakthrough is made in wireless charging, the wireless rechargeable sensor networks (WRSNs) are finally proposed. In order to reduce the charging completion time, most existing works use the “mo...As the technological breakthrough is made in wireless charging, the wireless rechargeable sensor networks (WRSNs) are finally proposed. In order to reduce the charging completion time, most existing works use the “mobilethen- charge” model—the Wireless charging vehicles (WCV) moves to the charging spot first and then charges nodes nearby. These works often aim to reduce the node’s movement delay or charging delay. However, the charging opportunities during the movement are overlooked in this model because WCV can charge nodes when it goes from one spot to the next. In order to use the charging opportunities, a speed grading method is proposed under the circumstance of variable WCV speed, which transformed the problem of final charging delay into a traveling salesman problem with speed grading. The problem was further solved by linear programming method. The simulation experiments show that, compared with the existing charging methods, the proposed method has a significant improvement in charging delay.展开更多
Multiple DAGs scheduling strategy is a critical factor affecting resource utilization and operating cost in the cloud computing. To solve the problem that multiple DAG scheduling cannot meet the resource utilization a...Multiple DAGs scheduling strategy is a critical factor affecting resource utilization and operating cost in the cloud computing. To solve the problem that multiple DAG scheduling cannot meet the resource utilization and reliability when multiple DAGs arrive at different time, the multiple DAGs scheduling problem can be transformed into a single DAG scheduling problem with limited resource available time period through multiple DAGs scheduling model based on backfill. On the basis of discussing the available time period description of resources and the sorting of task scheduling when the available time period is limited, the multiple DAGs scheduling strategy is proposed based on backfill. The experimental analysis shows that this strategy can effectively shorten the makespan and improve the resources utilization when multiple DAGs arrive at different time.展开更多
Pneumonia is one of the most common infectious diseases in clinical practice. In the field of pneumonia recognition, traditional algorithms have limitations in feature extraction and scope of application. To solve thi...Pneumonia is one of the most common infectious diseases in clinical practice. In the field of pneumonia recognition, traditional algorithms have limitations in feature extraction and scope of application. To solve this problem, a pneumonia recognition is proposed based on convolutional neural network. Firstly, the morphological preprocessing operation was performed on the chest X-ray. Secondly, the convolutional layer containing the 1 * 1 convolution kernel was used instead of a fully connected layer in the convolutional neural network to segment the lung field and obtain the segmentation. The index Dice coefficient can reach 0.948. Finally, a pneumonia recognition model based on convolutional neural network was established. The segmented images were trained and tested. The experimental results show that the average accuracy of the proposed method for pneumonia is up to 96.3%.展开更多
High-entropy effect is a novel design strategy to optimize properties and explore novel materials.In this work,(La_(1/5)Nd_(1/5)Sm_(1/5)Ho_(1/5)Y_(1/5))NbO_(4)(5RNO)high-entropy microwave dielectric ceramics were succ...High-entropy effect is a novel design strategy to optimize properties and explore novel materials.In this work,(La_(1/5)Nd_(1/5)Sm_(1/5)Ho_(1/5)Y_(1/5))NbO_(4)(5RNO)high-entropy microwave dielectric ceramics were successfully prepared in the sintering temperature(S.T.)range of 1210–1290℃via a solid-phase reaction route,and medium-entropy(La_(1/3)Nd_(1/3)Sm_(1/3))NbO_(4) and(La_(1/4)Nd_(1/4)Sm_(1/4)Ho_(1/4))NbO_(4)(3RNO and 4RNO)ceramics were compared.The effects of the entropy(S)on crystal structure,phase transition,and dielectric performance were evaluated.The entropy increase yields a significant increase in a phase transition temperature(from monoclinic fergusonite to tetragonal scheelite structure).Optimal microwave dielectric properties were achieved in the high-entropy ceramics(5RNO)at the sintering temperature of 1270℃for 4 h with a relative density of 98.2%and microwave dielectric properties of dielectric permittirity(ε_(r))=19.48,quality factor(Q×f)=47,770 GHz,and resonant frequency temperature coefficient(τ_(f))=–13.50 ppm/℃.This work opens an avenue for the exploration of novel microwave dielectric material and property optimization via entropy engineering.展开更多
The multi-objective optimization problem has been encountered in numerous fields such as high-speed train head shape design,overlapping community detection,power dispatch,and unmanned aerial vehicle formation.To addre...The multi-objective optimization problem has been encountered in numerous fields such as high-speed train head shape design,overlapping community detection,power dispatch,and unmanned aerial vehicle formation.To address such issues,current approaches focus mainly on problems with regular Pareto front rather than solving the irregular Pareto front.Considering this situation,we propose a many-objective evolutionary algorithm based on decomposition with dynamic resource allocation(Ma OEA/D-DRA)for irregular optimization.The proposed algorithm can dynamically allocate computing resources to different search areas according to different shapes of the problem’s Pareto front.An evolutionary population and an external archive are used in the search process,and information extracted from the external archive is used to guide the evolutionary population to different search regions.The evolutionary population evolves with the Tchebycheff approach to decompose a problem into several subproblems,and all the subproblems are optimized in a collaborative manner.The external archive is updated with the method of rithms using a variety of test problems with irregular Pareto front.Experimental results show that the proposed algorithèm out-p£performs these five algorithms with respect to convergence speed and diversity of population members.By comparison with the weighted-sum approach and penalty-based boundary intersection approach,there is an improvement in performance after integration of the Tchebycheff approach into the proposed algorithm.展开更多
Since traditional machine learning methods are sensitive to skewed distribution and do not consider the characteristics in multiclass imbalance problems,the skewed distribution of multiclass data poses a major challen...Since traditional machine learning methods are sensitive to skewed distribution and do not consider the characteristics in multiclass imbalance problems,the skewed distribution of multiclass data poses a major challenge to machine learning algorithms.To tackle such issues,we propose a new splitting criterion of the decision tree based on the one-against-all-based Hellinger distance(OAHD).Two crucial elements are included in OAHD.First,the one-against-all scheme is integrated into the process of computing the Hellinger distance in OAHD,thereby extending the Hellinger distance decision tree to cope with the multiclass imbalance problem.Second,for the multiclass imbalance problem,the distribution and the number of distinct classes are taken into account,and a modified Gini index is designed.Moreover,we give theoretical proofs for the properties of OAHD,including skew insensitivity and the ability to seek a purer node in the decision tree.Finally,we collect 20 public real-world imbalanced data sets from the Knowledge Extraction based on Evolutionary Learning(KEEL)repository and the University of California,Irvine(UCI)repository.Experimental and statistical results show that OAHD significantly improves the performance compared with the five other well-known decision trees in terms of Precision,F-measure,and multiclass area under the receiver operating characteristic curve(MAUC).Moreover,through statistical analysis,the Friedman and Nemenyi tests are used to prove the advantage of OAHD over the five other decision trees.展开更多
文摘In the market of agricultural products, the price of agricultural products is affected by production cost, market supply and other factors. In order to obtain the market information of agricultural products, the price fluctuation can be analyzed and predicted. A distributed big data software platform based on Hadoop, Hive and Spark is proposed to analyze and forecast agricultural price data. Firstly, Hadoop, Hive and Spark big data frameworks were built to store the data information of agricultural products crawled into MYSQL. Secondly, the information of agricultural products crawled from MYSQL was exported to a text file, uploaded to HDFS, and mapped to spark SQL database. The data was cleaned and improved by Holt-Winters (three times exponential smoothing method) model to predict the price of agricultural products in the future. The data cleaned by spark SQL was imported and predicted by improved Holt-Winters into MYSQL database. The technologies of pringMVC, Ajax and Echarts were used to visualize the data.
文摘Aiming at geological disaster monitoring and prevention work, a real-time monitoring and early warning system is proposed for low-power consumption of landslides to meet the needs of landslide monitoring in remote mountainous areas. The inclination angle of the mountain body was detected by a mechanical inclination sensor, and a plurality of inclination sensors were placed on each landslide body to form an array distribution. The landslide body was stereoscopically monitored. Each mesh node had a different node address, and different inclination thresholds were set in advance. When the sensor detection value reached the threshold, an alarm message was sent to the system main control end, and the main control end generated an audible and visual alarm, and an alarm message was sent at the same time. Compared with the current landslide warning system on the market, the system achieves expected results and its power consumption is extremely low. The sensor terminal of the mountain monitoring is powered by dry battery and can work for 6 years in the field without external power supply. It avoids the damages made by weather, livestock and human, and has broad application prospects.
基金supported by Guangxi key Laboratory Fund of Embedded Technology and Intelligent System under Grant No. 2018B-1the Natural Science Foundation for Distinguished Young Scholars of Jiangsu Province under Grant No. BK20160034+1 种基金the National Natural Science Foundation of China under Grant No. 61771488, No. 61671473 and No. 61631020in part by the Open Research Foundation of Science and Technology on Communication Networks Laboratory
文摘The high-frequency(HF) communication is one of essential communication methods for military and emergency application. However, the selection of communication frequency channel is always a difficult problem as the crowded spectrum, the time-varying channels, and the malicious intelligent jamming. The existing frequency hopping, automatic link establishment and some new anti-jamming technologies can not completely solve the above problems. In this article, we adopt deep reinforcement learning to solve this intractable challenge. First, the combination of the spectrum state and the channel gain state is defined as the complex environmental state, and the Markov characteristic of defined state is analyzed and proved. Then, considering that the spectrum state and channel gain state are heterogeneous information, a new deep Q network(DQN) framework is designed, which contains multiple sub-networks to process different kinds of information. Finally, aiming to improve the learning speed and efficiency, the optimization targets of corresponding sub-networks are reasonably designed, and a heterogeneous information fusion deep reinforcement learning(HIF-DRL) algorithm is designed for the specific frequency selection. Simulation results show that the proposed algorithm performs well in channel prediction, jamming avoidance and frequency channel selection.
基金supported by the National Science Foundation of China (Grant Nos.62267001,61906051)。
文摘With the popularity of online learning and due to the significant influence of emotion on the learning effect,more and more researches focus on emotion recognition in online learning.Most of the current research uses the comments of the learning platform or the learner’s expression for emotion recognition.The research data on other modalities are scarce.Most of the studies also ignore the impact of instructional videos on learners and the guidance of knowledge on data.Because of the need for other modal research data,we construct a synchronous multimodal data set for analyzing learners’emotional states in online learning scenarios.The data set recorded the eye movement data and photoplethysmography(PPG)signals of 68 subjects and the instructional video they watched.For the problem of ignoring the instructional videos on learners and ignoring the knowledge,a multimodal emotion recognition method in video learning based on knowledge enhancement is proposed.This method uses the knowledge-based features extracted from instructional videos,such as brightness,hue,saturation,the videos’clickthrough rate,and emotion generation time,to guide the emotion recognition process of physiological signals.This method uses Convolutional Neural Networks(CNN)and Long Short-Term Memory(LSTM)networks to extract deeper emotional representation and spatiotemporal information from shallow features.The model uses multi-head attention(MHA)mechanism to obtain critical information in the extracted deep features.Then,Temporal Convolutional Network(TCN)is used to learn the information in the deep features and knowledge-based features.Knowledge-based features are used to supplement and enhance the deep features of physiological signals.Finally,the fully connected layer is used for emotion recognition,and the recognition accuracy reaches 97.51%.Compared with two recent researches,the accuracy improved by 8.57%and 2.11%,respectively.On the four public data sets,our proposed method also achieves better results compared with the two recent researches.The experiment results show that the proposed multimodal emotion recognition method based on knowledge enhancement has good performance and robustness.
基金The National Natural Science Foundation of China (No.62262011)The Natural Science Foundation of Guangxi (No.2021JJA170130).
文摘Predicting the usage of container cloud resources has always been an important and challenging problem in improving the performance of cloud resource clusters.We proposed an integrated prediction method of stacking container cloud resources based on variational modal decomposition(VMD)-Permutation entropy(PE)and long short-term memory(LSTM)neural network to solve the prediction difficulties caused by the non-stationarity and volatility of resource data.The variational modal decomposition algorithm decomposes the time series data of cloud resources to obtain intrinsic mode function and residual components,which solves the signal decomposition algorithm’s end-effect and modal confusion problems.The permutation entropy is used to evaluate the complexity of the intrinsic mode function,and the reconstruction based on similar entropy and low complexity is used to reduce the difficulty of modeling.Finally,we use the LSTM and stacking fusion models to predict and superimpose;the stacking integration model integrates Gradient boosting regression(GBR),Kernel ridge regression(KRR),and Elastic net regression(ENet)as primary learners,and the secondary learner adopts the kernel ridge regression method with solid generalization ability.The Amazon public data set experiment shows that compared with Holt-winters,LSTM,and Neuralprophet models,we can see that the optimization range of multiple evaluation indicators is 0.338∼1.913,0.057∼0.940,0.000∼0.017 and 1.038∼8.481 in root means square error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE)and variance(VAR),showing its stability and better prediction accuracy.
基金supported by the National Natural Science Foundation of China under Grant 61762031the Science and Technology Major Project of Guangxi Province under Grant AA19046004+2 种基金the Natural Science Foundation of Guangxi under Grant 2021JJA170130the Innovation Project of Guangxi Graduate Education under Grant YCSW2022326the Research Project of Guangxi Philosophy and Social Science Planning under Grant 21FGL040。
文摘The recommendation system(RS)on the strength of Graph Neural Networks(GNN)perceives a user-item interaction graph after collecting all items the user has interacted with.Afterward the RS performs neighborhood aggregation on the graph to generate long-term preference representations for the user in quick succession.However,user preferences are dynamic.With the passage of time and some trend guidance,users may generate some short-term preferences,which are more likely to lead to user-item interactions.A GNN recommendation based on long-and short-term preference(LSGNN)is proposed to address the above problems.LSGNN consists of four modules,using a GNN combined with the attention mechanism to extract long-term preference features,using Bidirectional Encoder Representation from Transformers(BERT)and the attention mechanism combined with Bi-Directional Gated Recurrent Unit(Bi-GRU)to extract short-term preference features,using Convolutional Neural Network(CNN)combined with the attention mechanism to add title and description representations of items,finally inner-producing long-term and short-term preference features as well as features of items to achieve recommendations.In experiments conducted on five publicly available datasets from Amazon,LSGNN is superior to state-of-the-art personalized recommendation techniques.
文摘A five-axis camera stabilizer based on quaternion unscented Kalman filter algorithm is designed. It combined the unscented Kalman filter algorithm with the quaternion attitude solution and was solved by attitude sensor. By attitude algorithm, the motor in three directions of pitch, heading and roll in the stabilizer was accurately adjusted to control the movement of the three electronic arms. In order to improve the three-axis hand-held camera stabilizer’s performance, and to solve the jitter problem of up-and-down movement not being eliminated, two mechanical anti-shake arms were loaded under the stabilizer to balance the camera’s picture in pitch, roll, heading, and above and below five directions. Movement can maintain a stable effect. The simulation results show that the algorithm can effectively suppress the attitude angle divergence and improve the attitude calculation accuracy.
基金the National Natural Science Foundation of China (Grant No. 61762031)Guangxi Key Research and Development Plan (Gui Science AB17195029, Gui Science AB18126006)+3 种基金Guangxi key Laboratory Fund of Embedded Technology and Intelligent System, 2017 Innovation Project of Guangxi Graduate Education (No. YCSW2017156)2018 Innovation Project of Guangxi Graduate Education (No. YCSW2018157)Subsidies for the Project of Promoting the Ability of Young and Middleaged Scientific Research in Universities and Colleges of Guangxi (KY2016YB184)2016 Guilin Science and Technology Project (Gui Science 2016010202).
文摘Due to the development of E-Commerce, collaboration filtering (CF) recommendation algorithm becomes popular in recent years. It has some limitations such as cold start, data sparseness and low operation efficiency. In this paper, a CF recommendation algorithm is propose based on the latent factor model and improved spectral clustering (CFRALFMISC) to improve the forecasting precision. The latent factor model was firstly adopted to predict the missing score. Then, the cluster validity index was used to determine the number of clusters. Finally, the spectral clustering was improved by using the FCM algorithm to replace the K-means in the spectral clustering. The simulation results show that CFRALFMISC can effectively improve the recommendation precision compared with other algorithms.
基金The National Natural Science Foundation of China(No.62262011)the Natural Science Foundation of Guangxi(No.2021JJA170130).
文摘Fuzzy C-Means(FCM)is an effective and widely used clustering algorithm,but there are still some problems.considering the number of clusters must be determined manually,the local optimal solutions is easily influenced by the random selection of initial cluster centers,and the performance of Euclid distance in complex high-dimensional data is poor.To solve the above problems,the improved FCM clustering algorithm based on density Canopy and Manifold learning(DM-FCM)is proposed.First,a density Canopy algorithm based on improved local density is proposed to automatically deter-mine the number of clusters and initial cluster centers,which improves the self-adaptability and stability of the algorithm.Then,considering that high-dimensional data often present a nonlinear structure,the manifold learning method is applied to construct a manifold spatial structure,which preserves the global geometric properties of complex high-dimensional data and improves the clustering effect of the algorithm on complex high-dimensional datasets.Fowlkes-Mallows Index(FMI),the weighted average of homogeneity and completeness(V-measure),Adjusted Mutual Information(AMI),and Adjusted Rand Index(ARI)are used as performance measures of clustering algorithms.The experimental results show that the manifold learning method is the superior distance measure,and the algorithm improves the clustering accuracy and performs superiorly in the clustering of low-dimensional and complex high-dimensional data.
基金This research work was supported by the National Natural Science Foundation of China(Grant No.61762031)Guangxi Key Research and Development Plan(No.2017AB51024)Guangxi key Laboratory of Embedded Technology and Intelligent System,Guangxi Fundamental Laboratory for Embedded Technology and Intelligent Systems.
文摘Container virtual technology aims to provide program independence and resource sharing.The container enables flexible cloud service.Compared with traditional virtualization,traditional virtual machines have difficulty in resource and expense requirements.The container technology has the advantages of smaller size,faster migration,lower resource overhead,and higher utilization.Within container-based cloud environment,services can adopt multi-target nodes.This paper reports research results to improve the traditional trust model with consideration of cooperation effects.Cooperation trust means that in a container-based cloud environment,services can be divided into multiple containers for different container nodes.When multiple target nodes work for one service at the same time,these nodes are in a cooperation state.When multi-target nodes cooperate to complete the service,the target nodes evaluate each other.The calculation of cooperation trust evaluation is used to update the degree of comprehensive trust.Experimental simulation results show that the cooperation trust evaluation can help solving the trust problem in the container-based cloud environment and can improve the success rate of following cooperation.
基金The National Natural Science Foundation of China(No.61762031)The Science and Technology Major Project of Guangxi Province(NO.AA19046004)The Natural Science Foundation of Guangxi(No.2021JJA170130).
文摘There is a large amount of information in the network data that we canexploit. It is difficult for classical community detection algorithms to handle network data with sparse topology. Representation learning of network data is usually paired with clustering algorithms to solve the community detection problem.Meanwhile, there is always an unpredictable distribution of class clusters outputby graph representation learning. Therefore, we propose an improved densitypeak clustering algorithm (ILDPC) for the community detection problem, whichimproves the local density mechanism in the original algorithm and can betteraccommodate class clusters of different shapes. And we study the communitydetection in network data. The algorithm is paired with the benchmark modelGraph sample and aggregate (GraphSAGE) to show the adaptability of ILDPCfor community detection. The plotted decision diagram shows that the ILDPCalgorithm is more discriminative in selecting density peak points compared tothe original algorithm. Finally, the performance of K-means and other clusteringalgorithms on this benchmark model is compared, and the algorithm is proved tobe more suitable for community detection in sparse networks with the benchmarkmodel on the evaluation criterion F1-score. The sensitivity of the parameters ofthe ILDPC algorithm to the low-dimensional vector set output by the benchmarkmodel GraphSAGE is also analyzed.
文摘Pedestrian detection has a wide range of applications in daily life, and many fields require pedestrians to conduct detection with high precision and speed, which is an urgent problem to be solved. The traditional pedestrian detection method improves the detection performance by improving the classification algorithm and extracting more effective features. In this paper, a pedestrian detection method is proposed based on single shot multibox detector (SSD) model, which replaces the basic network part of SSD model with inception network structure with smaller parameters, faster running speed and stronger nonlinear expression ability. A high-performance network model for pedestrian detection was based on improved SSD. The experimental results show that the proposed method is faster than the original model, and the average precision of pedestrian recognition and location is 89.6%, which is 2.6% higher than the original model.
文摘As the technological breakthrough is made in wireless charging, the wireless rechargeable sensor networks (WRSNs) are finally proposed. In order to reduce the charging completion time, most existing works use the “mobilethen- charge” model—the Wireless charging vehicles (WCV) moves to the charging spot first and then charges nodes nearby. These works often aim to reduce the node’s movement delay or charging delay. However, the charging opportunities during the movement are overlooked in this model because WCV can charge nodes when it goes from one spot to the next. In order to use the charging opportunities, a speed grading method is proposed under the circumstance of variable WCV speed, which transformed the problem of final charging delay into a traveling salesman problem with speed grading. The problem was further solved by linear programming method. The simulation experiments show that, compared with the existing charging methods, the proposed method has a significant improvement in charging delay.
文摘Multiple DAGs scheduling strategy is a critical factor affecting resource utilization and operating cost in the cloud computing. To solve the problem that multiple DAG scheduling cannot meet the resource utilization and reliability when multiple DAGs arrive at different time, the multiple DAGs scheduling problem can be transformed into a single DAG scheduling problem with limited resource available time period through multiple DAGs scheduling model based on backfill. On the basis of discussing the available time period description of resources and the sorting of task scheduling when the available time period is limited, the multiple DAGs scheduling strategy is proposed based on backfill. The experimental analysis shows that this strategy can effectively shorten the makespan and improve the resources utilization when multiple DAGs arrive at different time.
文摘Pneumonia is one of the most common infectious diseases in clinical practice. In the field of pneumonia recognition, traditional algorithms have limitations in feature extraction and scope of application. To solve this problem, a pneumonia recognition is proposed based on convolutional neural network. Firstly, the morphological preprocessing operation was performed on the chest X-ray. Secondly, the convolutional layer containing the 1 * 1 convolution kernel was used instead of a fully connected layer in the convolutional neural network to segment the lung field and obtain the segmentation. The index Dice coefficient can reach 0.948. Finally, a pneumonia recognition model based on convolutional neural network was established. The segmented images were trained and tested. The experimental results show that the average accuracy of the proposed method for pneumonia is up to 96.3%.
基金Fundamental Research Funds of Shaanxi Key Laboratory of Artificially-Structured Functional Materials and Devices(No.AFMD-KFJJ-21210)the financial support from the National Natural Science Foundation of China(No.62061011)Guangxi Key Laboratory Fund of Embedded Technology and Intelligent System(No.2020-1-6).
文摘High-entropy effect is a novel design strategy to optimize properties and explore novel materials.In this work,(La_(1/5)Nd_(1/5)Sm_(1/5)Ho_(1/5)Y_(1/5))NbO_(4)(5RNO)high-entropy microwave dielectric ceramics were successfully prepared in the sintering temperature(S.T.)range of 1210–1290℃via a solid-phase reaction route,and medium-entropy(La_(1/3)Nd_(1/3)Sm_(1/3))NbO_(4) and(La_(1/4)Nd_(1/4)Sm_(1/4)Ho_(1/4))NbO_(4)(3RNO and 4RNO)ceramics were compared.The effects of the entropy(S)on crystal structure,phase transition,and dielectric performance were evaluated.The entropy increase yields a significant increase in a phase transition temperature(from monoclinic fergusonite to tetragonal scheelite structure).Optimal microwave dielectric properties were achieved in the high-entropy ceramics(5RNO)at the sintering temperature of 1270℃for 4 h with a relative density of 98.2%and microwave dielectric properties of dielectric permittirity(ε_(r))=19.48,quality factor(Q×f)=47,770 GHz,and resonant frequency temperature coefficient(τ_(f))=–13.50 ppm/℃.This work opens an avenue for the exploration of novel microwave dielectric material and property optimization via entropy engineering.
基金the National Natural Science Foundation of China(Nos.6156301261802085+5 种基金and 61203109)the Guangxi Natural Science Foundation of China(Nos.2014GhXN6SF AA1183712015GXNSFBA139260and 2020GXNSFAA159038)the Guangxi Key Laboratory of Embedded Technology and Intelligent System Foundation(No.2018A-04)the Guangxi Key Laboratory of Trusted Software Foundation(Nos.kx202011 and khx2601926)。
文摘The multi-objective optimization problem has been encountered in numerous fields such as high-speed train head shape design,overlapping community detection,power dispatch,and unmanned aerial vehicle formation.To address such issues,current approaches focus mainly on problems with regular Pareto front rather than solving the irregular Pareto front.Considering this situation,we propose a many-objective evolutionary algorithm based on decomposition with dynamic resource allocation(Ma OEA/D-DRA)for irregular optimization.The proposed algorithm can dynamically allocate computing resources to different search areas according to different shapes of the problem’s Pareto front.An evolutionary population and an external archive are used in the search process,and information extracted from the external archive is used to guide the evolutionary population to different search regions.The evolutionary population evolves with the Tchebycheff approach to decompose a problem into several subproblems,and all the subproblems are optimized in a collaborative manner.The external archive is updated with the method of rithms using a variety of test problems with irregular Pareto front.Experimental results show that the proposed algorithèm out-p£performs these five algorithms with respect to convergence speed and diversity of population members.By comparison with the weighted-sum approach and penalty-based boundary intersection approach,there is an improvement in performance after integration of the Tchebycheff approach into the proposed algorithm.
基金Project supported by the National Natural Science Foundation of China(Nos.61802085 and 61563012)the Guangxi Provincial Natural Science Foundation,China(Nos.2021GXNSFAA220074and 2020GXNSFAA159038)+1 种基金the Guangxi Key Laboratory of Embedded Technology and Intelligent System Foundation,China(No.2018A-04)the Guangxi Key Laboratory of Trusted Software Foundation,China(No.kx202011)。
文摘Since traditional machine learning methods are sensitive to skewed distribution and do not consider the characteristics in multiclass imbalance problems,the skewed distribution of multiclass data poses a major challenge to machine learning algorithms.To tackle such issues,we propose a new splitting criterion of the decision tree based on the one-against-all-based Hellinger distance(OAHD).Two crucial elements are included in OAHD.First,the one-against-all scheme is integrated into the process of computing the Hellinger distance in OAHD,thereby extending the Hellinger distance decision tree to cope with the multiclass imbalance problem.Second,for the multiclass imbalance problem,the distribution and the number of distinct classes are taken into account,and a modified Gini index is designed.Moreover,we give theoretical proofs for the properties of OAHD,including skew insensitivity and the ability to seek a purer node in the decision tree.Finally,we collect 20 public real-world imbalanced data sets from the Knowledge Extraction based on Evolutionary Learning(KEEL)repository and the University of California,Irvine(UCI)repository.Experimental and statistical results show that OAHD significantly improves the performance compared with the five other well-known decision trees in terms of Precision,F-measure,and multiclass area under the receiver operating characteristic curve(MAUC).Moreover,through statistical analysis,the Friedman and Nemenyi tests are used to prove the advantage of OAHD over the five other decision trees.