Machine learning approach for cost estimation in software project planning

Abstract

Successful organizing and handling of software projects depends extensively on accurate cost estimation. This study explores the effectiveness of machine learning models in estimating software project costs using datasets like Desharnais, Maxwell, and Kitchenham, aiming to prevent project delays and resource misallocation. It shows how model selection has a major impact on forecast accuracy through thorough assessment. An R-squared value (R2) of 0.804 indicates that the support vector machine (SVM) model performs exceptionally well in the Desharnais dataset. On the Maxwell dataset, linear regression (LR) stands out with a minimum mean absolute error (MAE) of 0.483 and the greatest R2 value of 0.607, while SVM has the lowest root mean squared error (RMSE) of 0.537. Similarly, on the Kitchenham dataset, LR and SVM are the top performers, with MAE of 0.201 and RMSE of 0.274, respectively, and R2 values of around 0.929. These findings highlight the importance of tailored model selection for accurate cost prediction, as LR and SVM continuously demonstrate reliability across varied datasets. ML techniques like LR and SVM can enhance software project planning and management by providing accurate cost estimation, with future research exploring ensemble learning and deep learning methodologies.