MLOps for Enhancing the Accuracy of Machine Learning Models using DevOps, Continuous Integration, and Continuous Deployment

Abstract

Machine learning (ML) integrated with development and operations (DevOps) is the key to solving the problem of deploying the latest machine learning models. This paper proposes one of the ways of integrating machine learning with DevOps. The need for this integration is endless as this provides seamless upgradation of the so-created models while also making managing and monitoring simple. The paper also provides light on practices of Continuous Integration/Continuous Deployment (CI/CD) and minimizing the unnecessary loss of time while training an ML model. The procedure followed includes CI/CD that contains jobs to train the models and to roll out the model with maximum performance. The main focus of this paper is the dynamic change of hyperparameters to achieve increased accuracy without the necessity of the physical presence of humans to change it. This research is independent of the type of machine learning model used and can be best followed for neural networks.