Software effort estimation requires the determination of one or more of the following estimates; effort (usually in person-months), project duration (in calendar time) and cost (in money). The ability to accurately estimate software project effort is essential for successful project planning, budgeting, and execution. This paper focuses on the development of an ensemble stacking model to enhance software effort estimation accuracy. The integration of a Kalman Filter (KFA) with various machine learning techniques, the model offers an improvement over traditional single-model approaches. Datasets of Albrecht, China, Cocomo81, Desharnais, Kemerer, and Maxwell were used for the model training and evaluation. Performance metrics of Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-Squared values were employed to validate the model. The results demonstrated a notable improvement in estimation accuracy, particularly in larger datasets, as compared to established models like the ensemble voting model. We made recommendations on the incorporation of additional datasets and hyper-parameter optimization to further enhance the model's performance.

Ensemble Stacking Effort Estimation Kalman Filter Software Project