数据挖掘下光纤通信网络异常动态检测方法OA

The channel capacity of fiber optic communication networks often approaches the Shannon limit,and large differences in network data lead to inaccurate sample partitioning and some missing feature components,and zero prior probability creates classification difficulties,so it is difficult to obtain accurate anomaly detection results.In view of the above,this paper proposes a dynamic detection method based on data mining.Establish an attribute feature set,design a classification objective function,and analyze data differences based on category standards,so as to complete the partitioning of feature categories.Introduce information entropy theory to compute weights and perform weighted data processing,so as to enhance clustering accuracy.Build a data set similarity relationship,integrate and process data with the same feature attributes,construct an integrated model,calculate error adjustment deviation to determine clustering centers,and export clustering results.Derive category probabilities based on Bayes' theorem and use Laplace smoothing algorithm to adjust prior probabilities to avoid zero probability.Set parameters and coefficients to mine partitioned sample data,analyze abnormal data to construct partitioned sample sets,cyclically evaluate and generate Bayesian classification probability results,and iteratively correct and output dynamic anomaly detection results.The experiments show that the proposed method has a high clustering PC value in fiber optic communication networks,with a maximum of 0.996,and a low PE value,with a maximum of only 0.016;the accuracy of anomaly detection reaches 100%,with a maximum false alarm rate of only 0.91%.The comprehensive advances of the proposed method in clustering accuracy,detection sensitivity,and false alarm control has been fully verified.It can be seen that the proposed method can provide reliable technical support for the safe and stable operation of fiber optic communication networks.