A network scene recognition algorithm that integrates temporal features
Due to different distribution of users and varying nature of business loads in changing network scenarios, the performance bottlenecks and common failure patterns in different network environments vary as well, such as dormitory coverage, classroom concurrency, and conference room interference. The network resource allocation schemes and actual operational standards also vary in different scenarios. This paper proposes a recognition algorithm framework based on semantic recognition improvement of tree model to address the network scene recognition problems. This framework employs semantic analysis to extract network temporal features and vertically express user activity trajectories. It is able to more intuitively characterize scenarios such as offices, dormitories, and canteens. Meanwhile, the model, installation density, and user load information of devices in the network environment are obtained by identifying the network environment characteristics fitted by the physical model of data communication, including path loss and interference between devices. Based on the decision tree model framework, the above two types of features are coupled to generate a scenario recognition algorithm framework. This global framework is characterized by intelligent recognition of network scenarios by coupling the traffic features and network planning features of the network environment. Our algorithm is validated on multiple datasets, demonstrating its effectiveness in identifying different scenarios.
actual scene analysistime series feature modelingscene recognitionfeature fusiontree model
国家科技期刊平台
NSTL
万方数据
