<![CDATA[Accurate effort estimation underpins on-time,on-budget software delivery. This study empirically assesses the baseline Constructive cost Model (COCOMO) by applying standard organic-mode parameters (a = 2.4, b = 1.05) to the COCOMONASA dataset, which contains 63 NASA projects ranging from 2 KLOC to 100 KLOC. Model ourputs are benchmarked against recorded person-month effort using Mean Absolute Error (MAE), Mean Magnitude of Relative Error (MMRE), and Predcitions at 25 percent error (PRED 0.25). Results show MAE values 295-661 person-months and an MMRE near 1.0, indicating average relative error of ~100 percent. PRED (0.25) equals 0.0, meaning no project is estimated within the industry-accepted 25% band. Sensitivity tests on 5- and 20-project subsets reveal similar patterns, confiriming that the inaccuracy is systemic rather than dataset-specific. Using uncalibrated COCOMO in present-day projects poses a high risk of severe under- or over allocation of resources, potentially trigerring budget overruns and schedule slips. By quantitatively exposing where and how the baseline model fails, this work provides a benchmark for and a roadmap toward-targeted parameter calibration and hybrid approaches that incorporate additional cost drivers or machine-learning techniques. Future research should explore automatic parameter tuning and context-aware hybrid models to achieve dependable effort estimation in contemporary software engineering.]]>
Copyrights © 2025
