Predictive Technique To Improve Classification On Continuous System Deployment

We introduce Last in First Focus (LIFF) approach for any system deployment. As per current market need most of the system deployment follow continues integration (CI) and continues deployment (CD) approach. In CI/CD, regular system release comes for deployment and it requires lot of post system deployment focus on last deployed or updated system. Continuous development, delivery and deployment is one of the essential parts of a product lifecycle. With addition of new features or major redesign with latest technology e.g. containerization, there may be sudden increase in tickets, which need to be handled effectively. When system does not perform expected behavior then system monitory module generates tickets. These tickets need to be analyzed and forward to corresponding department for resolution. With increase in number of tickets, more human resources may be required to perform this task. Here machine learning can provide helping hand in this job via analyzing the ticket and predicting the impacted system component for the resolution of ticket. The dataset collected after deployment of redesigned system component may be imbalanced due to more tickets on one system component only. This impacts the accuracy of classification metrics. This article resolved this problem by using Algorithmic level solution or Data level solution. Data level solution has achieved by oversampling of minority classes or under sampling of majority classes. Synthetic Minority Oversampling Technique, or SMOTE is used to address imbalanced dataset problem.