Docker and Kubernetes Pipeline for DevOps Software Defect Prediction with MLOps Approach

Software defects are common when it comes to software development. However, in reality, this is very detrimental for companies and organizations that are developing software. Prediction of software defects in the early stages of development can be a solution to this problem. Of course, the method used needs to be considered when developing a model for predicting software defects. The software continues to experience development, so the prediction model must always be updated so that it can adapt to existing conditions. This study proposes the MLOps approach, which combines development and operation processes to develop a software defect prediction model. We will create a prediction model and then create a Docker and Kubernetes pipeline to automate the entire software defect prediction process so that it can speed up the development process and have good performance. We are comparing the performance evaluation results of the proposed method with the traditional method, which is run manually by Docker. The results showed that the entire source dataset had a fairly good accuracy rate of 76%-83% and a good recall rate of 79%-94%. The precision and recall values were also very good. Apart from that, it also produces a good Fl-score value of 84%-90%. And the development time until the model’s release is shorter: the average time is 7:02 minutes. Performance monitoring on the built-in web server is also easy to do and shows very good results. The web server can receive up to 156. $6/$sec requests in all models based on the dataset used, with the highest error rate at 45.03%. The use of the Docker and Kubernetes pipelines with the MLOps approach has been proven to have good performance, and the development of software defect models can be sped up.