Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-05-10 Cooperative journals: 《计算机应用研究》
Abstract： With the rapid development of mobile Internet services, mobile applications such as augmented reality, virtual reality, and ultra clear video have become popular and IoT applications are emerging. The limited computing power and the lack of endurance of smart terminal devices are becoming more and more inadequate for these applications. Aiming at this situation, this paper proposes a computational offload strategy based on improved auction algorithm under the premise of combining intelligent device performance and server resources in the scenario of multi-user and multi-MEC server, the strategy consists of two important phases: the unloading decision-making phase. By considering the calculation task size, computing requirements and server computing power, network bandwidth and other factors, this paper proposes the basis for the uninstallation decision; In the task scheduling phase, by considering the time requirements of the computing task and the computing performance of the MEC server, this paper proposes a task scheduling model based on improved auction algorithm. The experiment proves that the computational offloading strategy proposed in this paper can effectively reduce the service delay, reduce the energy consumption of smart devices, and improve user satisfaction.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the quality control problem in crowdsourcing data processing, this paper proposed a weighted K-nearest neighbor voting method. This method not only considers the mark of a certain sample to return an answer, but rather obtains a more accurate answer by considering the neighbors of the sample comprehensively. At the same time, it applies appropriate weights to the neighbors of the sample to further improve the performance of the algorithm and maintain the simplicity of the traditional majority vote. The K-nearest neighbor vote can effectively solve the problem of lack of markup. By weighting the neighbors, it can solve the influence of the unbalanced mark. And the generalization of the algorithm is stronger. Through experiments in various situations, the results show that the proposed weighted K-nearest neighbor voting method has achieved good results.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： In order to improve the accuracy and efficiency of real time classification of remote sensing images, a real time classification algorithm of remote sensing images based on the ant colony optimization algorithm and independent feature sets is proposed. Firstly, wavelet features and color features of remote sensing images are Abstract: d, and the features form the feature vectors; then, the ant colony optimization is adopted to optimize the feature space, and the significant feature set of each class are selected independently to reduce the dimension of each feature sub-space; lastly, an independent extreme learning machine is trained for each class to realize the remote sensing images classification. Simulation experimental results based on the public remote sensing image dataset show that the proposed algorithm realizes a good classification accuracy and computational efficiency.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： To solve the problem of small sample of gait recognition in the field of counterterrorism and security issues, this paper proposed a novel gait virtual sample generation method based on deep CNN(Convolutional and Neural Network) and DLTL (Dual Learning and Transfer Learning) . Firstly, low-level of CNN model VGG19 extracted gait style feature map, and then it used the DL(Dual Learning) to carry on the style feature training. Thus it made style feature model. Moreover, high-level of VGG19 extracted gait context feature map, and then it used the TL (Transfer Learning) to make context feature map carry on the style characteristic learning. Finally, it obtained the virtual migration samples. Experimental results demonstrate that these virtual samples remained individual gait feature but style feature. So this method can effectively expand small sample size. At the same time, when the number of virtual samples increase to a certain number, gait recognition rate has improved. The method was compared with the existing virtual sample generation method. The result shows that the method has a better performance, which can generate virtual samples in large numbers and improve the recognition rate of gait recognition steadily.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-24 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the problem of ignoring the statistical significance of features in detection of Android malicious applications, an Android malicious application detection method based on statistical features was proposed. This method extracted the statistical characteristics of the training data set and used a clustering algorithm to preprocess the malicious data set for reducing the impact of individual differences on the experimental results. On the other hand, this method combined the features and various machine learning algorithms (such as linear regression, neural network, etc.) to establish a detection model. The accuracy rate of the two models proposed by this method could reach more than 95%, and the detection time could be greatly reduced compared with the comparison experiment. Experimental results show that the statistical characteristics of the application can be used to distinguish between benign and malicious applications, and preprocessing the data by clustering algorithm can improve the detection accuracy.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-12 Cooperative journals: 《计算机应用研究》
Abstract： In view of the characteristic of multi-dimensional and multi-layer in traffic accident data, this paper proposed a new model to research the main reasons and potential rules in traffic accidents. The model starts from the four main dimensions such as the drivers, the vehicles, the time-address and the environment, and uses a way which based on AHP and hybrid Apriori-Gentic algorithm to mine causes of accident. First of all, the AHP sorted the importance of the influencing factors about accident. Then on the basis of objective analysis, the model quantified the influencing factors and selected the main influencing factors. Finally the model combined the genetic algorithm with the Apriori to directional analyze the main influencing factors and find the association rules out. The experimental result shows that the model could reduces the generation of useless rules and improves the accuracy of mining, which has certain scientific significance and application value.
Hits
Hits
Downloads
Comment