Predictive analysis of incidents based on software deployments

Abstract

A high number of IT organizations have problems when deploying their services, this alongside with the high number of services that organizations have daily, makes Incident Management (IM) process quite demanding. An effective IM system need to enable decision makers to detect problems easily otherwise the organizations can face unscheduled system downtime and/or unplanned costs. By predicting these problems, the decision makers can better allocate resources and mitigate costs. Therefore, this research aims to help predicting those problems by looking at the history of past deployments and incident ticket creation and relate them by using machine learning algorithms to predict the number of incidents of a certain deployment. This research aims to analyze the results with the most used algorithms found in the literature.