一种融合时序特征的网络场景识别算法

A network scene recognition algorithm that integrates temporal features

针对网络场景识别问题,提出了一种基于语义识别改进的树模型识别算法框架,利用语义分析提取网络时序特征,纵向表达用户活动轨迹,可更加直观地表征办公、宿舍和食堂等场景。同时,通过对数据通信物理模型拟合的网络环境特征(设备之间的路损、干扰等)进行识别,获得网络环境内设备的款型、安装的疏密和用户的负载等信息。进而基于决策树模型框架,耦合上述两类特征生成场景识别算法框架。通过耦合网络环境的流量特征及网络规划特征,全局框架具有智能识别网络场景的特点。算法针对多个数据集样本进行验证,证明所提方案均能对不同场景进行有效识别。

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.