Enhancing Software Cost Estimation Using Feature Selection and Machine Learning Techniques

Software cost estimation is a crucial aspect of software project management, significantly impacting productivity and planning. This research investigates the impact of various feature selection techniques on software cost estimation accuracy using the CoCoMo NASA dataset, which comprises data from 93 unique software projects with 24 attributes. By applying multiple machine learning algorithms alongside three feature selection methods, this study aims to reduce data redundancy and enhance model accuracy. Our findings reveal that the principal component analysis (PCA)-based feature selection technique achieved the highest performance, underscoring the importance of optimal feature selection in improving software cost estimation accuracy. It is demonstrated that our proposed method outperforms the existing method while achieving the highest precision, accuracy, and recall rates.