Enhancing Software Effort Estimation with Machine and Deep Learning Strategies
Vifert Jenuben Daniel., M Rithani · 2025
The software engineering area is frequently confronted with the difficulty of decision making under uncertainty, particularly in the estimation of software effort, which is critical in project planning and management. This work presents novel prediction models that utilize machine learning and deep learning techniques to enhance the accuracy of software effort estimation. The machine learning suite, comprising Linear Regression, K-Nearest Neighbors (k=3), and Support Vector Machines with linear kernels, demonstrated a promising model score, with Linear Regression achieving 76.8%. The deep learning ensemble, including Simple RNN, LSTM, and FNN, was also evaluated using the Desharnais dataset. Among these, the LSTM model outperformed with an R-squared score of 80.4%, showcasing its superior ability to capture and reflect the underlying patterns in project data. These experimental results underline the effectiveness of AI-based models, affirming their potential in improving software effort estimation with notable precision.