Improvement and Optimization of Machine Learning Algorithms based on Intelligent Computing

Abstract

In this paper, a sensor cloud data intrusion detection framework is proposed. The framework uses parallel discrete optimization techniques for feature refining and incorporates machine learning principles to improve sensing cloud security. Firstly, a set of optimal feature evaluation criteria is established, and a parallel discrete optimization feature extraction system is built to reduce the data dimension and strengthen the stability of feature processing. Then, a widely used discrete optimization algorithm is developed, proving its global convergence. The optimal feature set is obtained through parallel screening feature subsets. Finally, using these features and distributed fuzzy cluster analysis, the intrusion behavior of the sensing cloud is accurately detected. This method incorporates the concept of intelligent iterative evolution and self-regulating clustering strategy, which not only overcomes the local optimal trap that the conventional fuzzy clustering algorithm may encounter but also realizes the automatic adjustment of the number of clusters. The experimental data show that the intrusion detection algorithm performs excellently in providing accurate intrusion determination results. Compared with other detection algorithms, the accuracy of anomaly detection and the reduction of missing detection rate is significantly improved. In addition, the algorithm shows anti-interference solid ability and can maintain stable performance in noisy environments.