Detection of Errors in Websites Using CI/CD Tools and Automation with Machine Learning Techniques
AbstractErrors in web-based applications negatively affect user experience and cause operational efficiency losses for businesses. This study proposes a hybrid error management system integrated into CI/CD (Continuous Integration and Continuous Deployment) pipelines. The system consists of two main components: real-time monitoring of system metrics (API response codes, CPU, memory, disk usage, etc.) using Prometheus and Grafana, and analysis of user logs with ELK Stack (Elasticsearch, Logstash, Kibana). The collected data was cleaned in the preprocessing stage and balanced using the SMOTE-ENN method. For error classification, a rule-based model (if-else) was compared with machine learning algorithms (SVM, KNN). The system was containerized with Docker and Kubernetes and deployed via Jenkins, reducing error detection time. This study provides a framework combining rule-based and machine learning approaches to optimize traditional error management processes.