Intelligent alarm data compression algorithm for data network operation and maintenance
FAN Ming
Abstract：[Objective]With the explosive growth of network scale and complexity,traditional network operation and maintenance(O&M)technologies face challenges such as low accuracy,severe noise interference,and difficulties in root cause localization when processing massive alarm data.Existing association rule algorithms such as Apriori and FP-Growth,fail to deeply analyze the correlations and hierarchical root causes in alarm data,resulting in low compression efficiency and high false positive rates.This paper aims to design an intelligent alarm data compression algorithm by introducing a graph convolutional network(GCN)to address the limitations of traditional methods in mining data correlations,suppressing noise,and analyzing multi-level root causes,thus enhancing the intelligence level of network O&M and improving alarm processing efficiency.[Methods]To tackle the heterogeneity and redundancy of alarm data,a dynamic preprocessing mechanism based on sliding time windows was proposed.This mechanism constructed a high-precision alarm transaction database through time synchronization rules and redundancy removal operations.Next,the preprocessed alarm sequences were transformed into graph-structured data,where node feature matrices and adjacency matrices characterized alarm events and their correlations.A multi-layer GCN model was then designed:local convolution aggregated neighborhood node features,normalization techniques resolved structural imbalance in the graph data,and ReLU activation functions enhanced nonlinear feature extraction.Key parameter configurations included input feature dimensions,hidden layer structures,the Adam optimizer,and Dropout mechanisms,all balancing model complexity with generalization.Finally,performance differences between GCN and ResNet,Apriori,and FP-Growth were compared.[Results]Experimental results demonstrate that the proposed algorithm significantly outperforms traditional methods in both accuracy and runtime.Specifically,when the data volume reaches 6000 entries,GCN achieves stable alarm accuracy exceeding 92%,surpassing Apriori(83%),FP-Growth(87%),and ResNet(84%—94%)with smaller fluctuations.In terms of efficiency,GCN's average processing time is comparable to FP-Growth(with a difference of less than 5%for data volumes exceeding 1000 entries)and significantly lower than ResNet.Moreover,by capturing nonlinear correlations and hierarchical root causes in alarm data,GCN effectively suppresses the noise interference,validating its robustness in complex network environments.[Conclusions]The proposed GCN-based alarm compression algorithm achieves high-precision,low-redundancy alarm information extraction by deeply integrating spatiotemporal features and topological correlations of alarm data.Compared to traditional methods,it shows significant advantages in both accuracy and efficiency,providing reliable technical support for intelligent network O&M.Future work will focus on lightweight model design and real-time optimization to further adapt to large-scale dynamic network scenarios.
Keywords：network alarmdata redundancyfalse alarmdata compressionApriori algorithmFP-Growth algorithmroot cause analysisassociation relationship
Publication Date：2026-01-25
Online Publishing Date：2026-03-17(First online date of this platform, not the publication date of the document)
Pages：8( 46-53 )