Article Preview
Top1. Introduction
The software industry has become one of the fastest-growing industries. However, software development is still a high-risk sector that entails many risks (Hu et al., 2013). Risk management has become one of the main activities of software development because the failures in the software project are mostly due to poor and inefficient risk control. Comparing broad assignments that have effectively surpassed their costs as well as scheduling quotes against those that have been postponed. Over or even eventually reached their budgets, half a dozen common issues, poor calculations of costs, (Patil & Ade, 2014) analysis of requirements, and risk management have more influence than this occurs at the end of the software development life cycle (SDLC) and failure at this stage may increase maintenance cost in some cases, redevelopment chances opportunities would be much more significant, and poor requirement analysis and risk management have more effect as this happens at the end of the SDLC (Hu et al., 2015). The specifications of software engineering are the key and significant factor of software development. You cannot develop software best and perfectly according to users’ requirements until you did not complete proper risk management. The survey showed that the success rate was only 37 percent for the Standish group. Mainly 65 percent was due to software failure problems (Catal, 2011). In software development, risk prediction is obligatory if project success is to be recognized, classified, and prioritized earlier. The demand collection phase is the most significant and challenging step in SDLC. At this point, the risks must be addressed and saved for future projects (Shaukat et al., 2018). Consequently, risks in software requirements must first be forecast so that projects can succeed. Since risk assessment will be more highly beneficial at this stage and increase software productivity. It also helps to reduce the chances of software failure if the risk is properly managed at this stage. Frequent solutions for software risk prediction are available in various stages in SDLC, while infrequent methods are provided to predict risk in past studies (Usman et al., 2014). The very first stage of the software risk prediction model is risk identification, where the risk/project manager will identify the requirements traditionally. The requirements from software requirement specification (SRS) having risk threats were marked and checked for further analysis. After the checklist is completed headed to the next stage (Xu et al., 2000). However, in this research, we applied different Machine Learning (ML) techniques to find out the best solution for risk prediction in software requirements. These techniques include Support Vector Machine (SVM), Logistic Regression (LR), Naïve Bayes (NB), Random Forest (RF), Stochastic Gradient Descent (SGD), Gaussian NB (GNB), K-Nearest Neighbor (KNN), Linear Decrement Analysis (LDA), and Decision Tree (DT) that are evaluated using several assessments measure e.g., precision, F-measure, Matthew’s Correlation Coefficient (MCC), recall and accuracy.
Hereinafter, the paper is organized as Section 2 presents the literature review of the existing model. Section 3 details the entire research methodology and dataset with employed ML techniques. Section 4 discusses the results, analysis, and discussion. Section 5 presents the threats to validity that may happen due to certain reasons and finally, Section 6 concluded the paper with the overall outcome and recommendations.