Command information system state monitoring method and system, medium, equipment and terminal

Abstract

The invention belongs to the technical field of command information systems, and discloses a command information system state monitoring method and system, a medium, equipment and a terminal, and themethod comprises the steps: carrying out SVM pre-training through a sample set; utilizing the support vector to generate a potential new sample near the pre-classification surface; through the thoughtof zoning oversampling, zoning is carried out by taking the classification hyperplane as the center according to the distance, and new samples uniformly distributed in the current band are generatedband by band so as to keep the balance of the number of various samples. According to the method, the historical monitoring samples are subjected to oversampling processing, so that the influence caused by the imbalance of the number of positive / abnormal state samples is eliminated; by correcting the triggering condition of model updating and introducing a sample retention and forgetting mechanism, the training number of samples in the incremental learning process is reduced so as to reduce the time overhead. Experimental results prove that the classification effect in a real system state sample set is achieved, and compared with an existing algorithm in different unbalanced data sets, the capacity of dealing with unbalanced data in the incremental learning process is fully shown.

Description

technical field [0001] The invention belongs to the technical field of command information systems, and in particular relates to a state monitoring method, system, medium, equipment and terminal of a command information system. Background technique [0002] At present: For the monitoring of system status, many cloud service providers have opened resource running status data as a key service to users, and developed corresponding resource monitoring tools in cloud environments, such as Amazon's CloudWatch, Microsoft's Azure Monitor and so on. These tools analyze the running status of the user's subscription service and provide early warning for the user by monitoring and processing indicators such as CPU / RAM utilization, network traffic, and response time. In this context, a lot of research has emerged in the civilian field on the monitoring of system status. Use the external MAPE-K (Monitor, Analyze, Plan, and Execute, with a Knowledge Base common to all components) loop to...

AI Technical Summary

Problems solved by technology

The smaller the value of C, the greater the tolerance for the violation of the constraints; if C→+∞, the samples that violate the constraints cannot be tolerated, and the optimization problem 2 degenerates into the optimization problem 1

[0028] (1) For the problem of system status monitoring, the existing work deals with it by reducing or increasing the corresponding samples, resulting in the decline of the classification effect of the model; while the classic oversampling method has a certain blindness during oversampling, which easily leads to the synthesis of new samples are of poor quality

[0029] (2) Merging the new sample set with the original sample set will bring a large computational overhead, especially in the online learning scenario, retraining every time will seriously affect the real-time performance of learning; in addition, with the new As the sample set continues to increase, the cost of re-learning will also increase, which will have an adverse effect on the performance of the classifier.

Images