Exoloiting context- aware data and elastic analytics techniques for fault detection on large and complex communication systems

Abstract

Nowadays, cloud computing systems, software defined networks, content delivery networks, which contain numerous servers and network devices, are growing increasingly. Therefore, there is a big challenging to manage the scalability, availability, heterogeneity and importance of those system, whereas the human ability and assistant software are restricted. There is continuously a request of creating new methods and tools that support administrators in monitoring errors. Although, analysis event monitoring and fault detection has been considered by several approaches. However, this thesis points to study and apply Random Forest technique to analyze fault dataset. In fact, most of the existing approach usually points to detect and solve error on log event, message and trace by administrator's expertise. In fact, these methods intensely depend on human being. So, applying analytic techniques to determine significant facts of a problem, which helps administrator deal with the error and then reducing the dependency of human being. In computer science, there are various analytic techniques for classification and regression such as Decision Tree, K-means and Random Forest, which is one of them. Although, this technique is younger than Neural Network, Decision Tree, Kmeans, or Bayesian and existing some weakness; however, it has various strong features in classification and regression. In this study, Random Forest technique applied to estimate fault particularly in bug report. Last but not least, this proposal is contributing apart from the content of the publishing research paper namely “Analysis of Software Bug Reports Using Random Forest [1]” on ACIIDS 2018 conference