Research On Anomaly Detection Technology Of Optical Wireless Converged Network Based On Deep Learning

With the continuous development of communication technology and intelligent terminals,in order to meet the needs of emerging services such as smart cities,metaverses,and air-space-ground integration network,optical-wireless converged networks are more and more widely used.In the actual network operation and maintenance process,optical-wireless converged network anomaly detection can effectively avoid problems such as network conflicts and network crashes,thereby ensuring stable quality of service provided by optical and wireless converged networks.The existing mainstream optical wireless network anomaly detection methods mainly rely on preset threshold systems and manual identification and reasoning.They have not realized large-scale automation and intelligent detection,and cannot predict the occurrence time of network anomalies in advance.In addition,there is no unified solution in the existing technical solutions for whether there are real abnormal labels in the optical-wireless converged network data set.In order to solve the above problems,it is necessary to realize the intelligent and systematic anomaly detection of the optical-wireless converged network.This paper introduces a deep learning algorithm,conducts research on improving the accuracy of optical network index prediction and optical network anomaly detection,and proposes an optical wireless network anomaly detection architecture based on deep learning.The main contents are as follows:Firstly,for the more complex structure of optical-wireless converged networks under emerging services,this paper proposes an optical wireless network anomaly detection framework based on deep learning algorithms,and uses statistical correlation analysis algorithms to find out massive network indicators that are strongly related to abnormal factors.At the same time,a long-short-term memory neural network prediction model is built to predict the future time series data of optical wireless network indicators.The simulation results show that on the real optical-wireless converged network data set,the model can accurately predict the future time series values of optical network indicators,and the prediction accuracy can reach 93%.Secondly,for the optical wireless network nodes with real abnormal labels,this paper analyzes the strong correlation indicators that cause abnormal factors,and uses the deep neural network algorithm to build a supervised optical wireless network anomaly detection and classification model.The future time series data obtained in the previous step On the basis of accurate classification,it avoids the process of new data recalculation in the traditional clustering model.At the same time,the simulation results show that on the real optical-wireless converged network data set,the classification model can complete the classification of abnormal data with an accuracy rate of 98%,which is 13%higher than the traditional machine learning classification algorithm.Thirdly,the optical wireless network anomaly detection architecture based on the deep learning algorithm proposed in this paper also considers the situation of network nodes lacking real anomaly labels.This architecture uses a density-based clustering algorithm to build an unsupervised optical wireless network anomaly detection clustering model.Appropriate training standards and evaluation systems are established,and the task of anomaly detection in optical wireless networks that lacks real anomaly labels is effectively completed.In summary,the optical-wireless converged network anomaly detection architecture based on deep learning proposed in this paper has high applicability.It realizes the automation and intelligence of optical wireless network anomaly detection.It also improves the stability and robustness of optical-wireless converged network in emerging network service scenarios.