Software Cost Estimation: A Comparative Analysis Of Traditional And Machine Learning Approaches

Abstract

Estimating software costs is essential to project management because it helps businesses allocate resources efficiently. Despite their widespread use, traditional models like COCOMO frequently have drawbacks because they rely on assumptions and predefined parameters that are not well suited to contemporary software development techniques. Machine learning-based models, on the other hand, provide a data-driven strategy by utilizing past project data to increase estimation accuracy. This study compares and contrasts contemporary machine learning methods with conventional cost estimation models. We go over the drawbacks of traditional methods, investigate sophisticated regression strategies, and suggest a hybrid model that combines neural networks and XGBoost for increased accuracy. The superiority of our method in reducing estimation errors is shown by empirical results. The findings show that machine learning-based cost estimation can significantly enhance software engineering decision-making and resource planning. Neural networks, machine learning, XGBoost, COCOMO model, software cost estimation, and predictive analytics are some examples of index terms.