Software Effort Estimation Using Stacked Ensemble Technique and Hybrid Principal Component Regression and Multivariate Adaptive Regression Splines

In today’s rapidly evolving digital market, the demand for faster and more advanced software products requires accurate, cost-effective, and timely solutions. Precise software effort estimation is crucial for successful development. However, inaccurate estimations using various techniques have led to project complications. The research explores stacked ensemble techniques, combining base ensembled regression methods like Decision Tree, principal components regression (PCR), Random Forest, NeuralNet, glmnet, XGBoost, Earth, and Support Vector Machine. PCR emerged as the most effective technique. To address longer computation times in ensemble approaches, a hybrid model incorporating Principal Component Regression and multivariate adaptive regression splines (earth) was proposed, yielding accurate predictions with minimized computing time. Evaluation metrics included Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Squared Error (MSE), PRED (25), R-squared coefficients, and Computation Time. Implementing the hybrid approach aims to optimize effort estimation accuracy and reduce computational time for more effective software project management, ultimately leading to enhanced project outcomes and increased customer satisfaction.