UNSUPERVISED MACHINE LEARNING FOR FEEDBACK LOOP PROCESSING IN COGNITIVE DEVOPS SETTINGS

Abstract

Abstract—Agile development and quick adaptation to new needs are two of the most pressing issues facing software applications and systems today. The need for separate deployments, such as in DevOps, arises from this. A system's quality, particularly when it is continually built, is greatly affected by continuous monitoring and feedback creation, which are essential components of DevOps because of the short release cycles and flexibility they include. To achieve this goal, we offer a system for feedback that integrates data from operations and development. This system can identify patterns, spot unusual behavior, and feed that information back into development, providing a more thorough investigation into production anomalies. In order to achieve this goal, we describe the dataset using two unsupervised machine learning approaches, k-means clustering and archetypal analysis. Based on the findings, we classify new data points as normal or abnormal. An application is being built inside a big industrial organization that provides real-time data for testing and evaluation purposes. This application is designed to support the feedback loop, which consists of continuous development, planning, deployment, and monitoring.