A Classification and Identification Method for Overload of Network Big Data Based on Rels-Tsvm

Abnormal network loads lead to network congestion, system performance degradation, or service unavailability. In order to improve the quality of network operation, a classification and identification method for overload of network big data based on RELS-TSVM was proposed. Based on network big data samples, a batch processing learning algorithm is used to transform the problem of extracting abnormal load features from network big data into an optimization problem. The gradient descent algorithm is used to optimize the feature extraction objective function and complete the extraction of abnormal load features. After extracting the abnormal load features of network big data, they are input into the RELS-TSVM model, an abnormal load classification and recognition function is constructed, and the classification and recognition of network big data loads is completed. The experimental results show that compared with existing methods for abnormal load classification and recognition, the accuracy of this method is significantly improved, the maximum accuracy reached 0.98. In addition, this method can reduce the traffic consumption in the process of overload classification and recognition, and has superior practical application performance.