Metaheuristic-based cost-effective predictive modeling for DevOps project success

Over the decade, DevOps practices have gained popularity within software development organizations for managing the dynamic behavior of system development. Implementing DevOps introduces risks and challenges that increase the difficulties in software development activities, which consequently leads to project failures. This study presents a probability-based predictive model to estimate the success or failure of DevOps projects based on the 13 most significant features identified from literature and data gathered from the DevOps practitioners using a survey questionnaire. The Naïve Bayes Classifier and Logistic Regression (LR) models allied with Grey Wolf Optimization (GWO) have been used to measure the efficiency with the cost of implementing DevOps practices. The results of the study highlighted that NBC with GWO increased the success probability from 0.4954 to 0.9971, with the cost rising from 0.2577 to 0.5000. Similarly, LR with GWO also presented an increase in success probability from 0.2880 to 0.9839, along with an increase in cost from 0.2423 to 0.3558. In conclusion, the developed prediction model based on identified features could help DevOps software development practitioners to implement DevOps projects cost-effectively and successfully.