The initial stage of the software development lifecycle is the requirement gathering and analysis phase. Predicting risk at this phase is very crucial because cost and efforts can be saved while improving the quality and efficiency of the software to be developed. The datasets for software requirements risk prediction have been adopted in this paper to predict the risk levels across the software projects and to ascertain the attributes that contribute to the recognized risk in the software projects. A supervised machine learning technique was used to predict the risk across the projects using Naïve Bayes Classifier technique. The model was able to predict the risks across the projects and the performance metrics of the risk attributes were evaluated. The model predicted four (4) as Catastrophic, eleven (11) as High, eighteen (18) as Moderate, thirty-three (33) as Low and seven (7) as insignificant. The overall confusion matrix statistics on the risk levels prediction by the model had accuracy to be 98% with confidence interval (CI) of 95% and Kappa 97%.

Keywords : Naïve Bayes Classifier, SDLC, Risk Prediction, Software Projects, Risk Outcomes, Risk Levels