Prediction of Risk Percentage in Software Projects by Training Machine Learning Classifiers

Abstract

Recently, software project failures have been increasing due to lack of planning and budget constraints. In this regard, identifying the suitable software model with the consideration of risk factors is imperative. Therefore, this study investigates the key software models utilized in the industry through an interaction with software development experts and literature survey. In this study, 15 standard indicators were chosen where a survey was conducted through a questionnaire. The major performance metrics which were taken into consideration are network, security, software, machine learning, internet of things and application programming interface. We proposed a novel framework for the received dataset through questionnaire in which the machine learning classifiers were applied and risk predictions for each of the identified software models were accomplished. Using this outcome, software product managers can identify the appropriate software model according to the software requirements along with risk prediction percentage.