Enhancing continuous integration predictions: a hybrid LSTM-GRU deep learning framework with evolved DBSO algorithm

• Published in: Computing 2025, Vol.107 (1), p.9, Article 9
• Main Authors: Benjamin, Jetty, Mathew, Juby
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial Intelligence; Computer Appl. in Administrative Data Processing; Computer Communication Networks; Computer Science; Deep learning; Industrial development; Information Systems Applications (incl.Internet); Machine learning; Optimization; Regular Paper; Software development; Software Engineering; Success; Time dependence; Time series; Version control
• ISSN: 0010-485X; 1436-5057
• DOI: 10.1007/s00607-024-01370-2

DevOps, an advanced software engineering methodology widely adopted in the software development industry, facilitates the rapid release of features and versions to production environments. The term "DevOps" is a fusion of "Development" and "Operations." It is implemented through a series of processes known as Continuous Integration and Continuous Delivery, which form the core of DevOps operations. Continuous Integration involves the seamless integration of small code increments into the version control system. This practice initiates a continuous integration environment where ongoing code changes are merged and tested. To optimize this environment, the study and analysis of related metrics are essential. The primary goal of this research is to identify interdependent key metrics that significantly influence the outcomes of builds within a continuous integration environment. This study introduces four interdependent metrics relevant to the continuous integration environment, positively impacting its effectiveness. A set of rules is derived, and a hypothesis is formulated, substantiating the interdependence of these metrics. To validate the metric rules and their association with build success a hybrid LSTM-GRU model is employed. CI build outcome data exhibits temporal dependencies and a sequential nature, making it suitable for time series analysis. Leveraging time series techniques can provide valuable insights into the dynamics of the CI process, enabling better decision-making and optimization of software development workflows. So we introduce hybrid LSTM-GRU model for predicting the build outcome and validating the DKMR. The analysis underscores that metrics such as TSC (Time between Successive Commits), BBFT (Build Breakage Fixing Time), LC (Long Commit), and BT (Build Time) are mutually dependent and have a constructive influence on build outcomes. In culmination, an algorithm called the "Dynamic Build Success Optimization Algorithm," based on these interdependent key metrics, has been developed. This algorithm is designed to enhance the efficiency and reliability of the continuous integration environment, contributing to the overall success of DevOps practices.