Detection of Faulty node with Hybrid Machine Learning using SVM model

The huge volume of data generated these days from a wide variety of sources has contributed to the widespread use of distributed data processing methods. In today's sophisticated huge computing systems, occupations are broken down into multiple smaller tasks that are carried out in parallel in order to improve work learning results and minimise consumption. Nevertheless, responding with straggler processes, which are processes that run slowly and contribute to an increase in the overall response time, is a common performance problem in such platforms. In this research, we suggested utilizing a hybrid machine learning approach to find struggler nodes in a large distributed environment. These nodes would be identified by their inability to successfully complete tasks. In this paper, we proposed a hybrid machine learning model for detecting faulty nodes in large distributed machines with collaboration of reinforcement and supervised machine learning. The large Virtual Machine (VM's) log data has been collected from the distributed environment and proceeded with reinforcement learning algorithm for module training and supervised machine learning for module testing. According to extracted features, reinforcement learning encompasses an activation function that generates the label for the respective node, whether healthy or weak. In the testing phase, the natural world VM's log data has been collected and evaluated with supervised machine learning classifiers. Several machine learning classification algorithms have evaluated and acquired the results. The SVM provides higher accuracy over the other machine learning classifiers with our reinforcement learning model.