FEPP: IMPROVING SOFTWARE REQUIREMENTS RISK PREDICTION THROUGH HYBRID RULE EXTRACTION AND MULTI-CLASS LEARNING

Abstract

Predicting risks in software requirements, a
critical and vital component of the Software
Development Life Cycle (SDLC), is challenging
due to the growing complexity of software
projects. An inability to properly predict such
risks may result in the failure of a software
project. Risk prediction is more important in
software requirements as it is the first phase of
any software project. Therefore, this study
proposes a unique approach for risk prediction in
software requirements called ForExPlusPlus
(FEPP). The proposed model is benchmarked
against standard models including K-nearest
Neighbour (KNN), Naïve Bayes (NB), Logistic
Model Tree (LMT), Random Forest (RF), and
Support Vector Machine (SVM). These models
are trained using the Zenodo repository dataset,
and the outcomes are assessed using common
evaluation standards. The precision, F-measure
(FM), and Mathew's correlation coefficient
(MCC) are used to critically evaluate the
models' accuracy analysis. The Kappa Statistic
(KS) and Mean Absolute Error (MAE) are used
to evaluate the error rate. With an accuracy of
96.84%, the recommended FEPP outperforms
KNN, which has the lowest accuracy of 50.99%.