A Hybrid Machine Learning Approach for Software Project Effort and Cost Estimation

This paper presents a hybrid machine learning approach for software project effort and cost estimation. Accurate estimation of software development effort is critical for project planning, budgeting, and resource allocation. However, traditional estimation techniques such as COCOMO and Function Point Analysis often struggle to handle the complexity of modern software development environments. The proposed system integrates machine learning with traditional estimation methods to improve prediction accuracy and practical usability. A Random Forest classifier is used to predict the expected team experience level, while an analogous estimation approach is applied to estimate project cost based on historical project data and adjusted function points. The system is implemented as a complete end-to-end solution consisting of a FastAPI backend, a machine learning model developed using scikit-learn, SQLite database storage, and a web-based interface for generating predictions. The architecture enables real-time estimation, logging of prediction requests, and efficient management of project data. Experimental results show that the hybrid model achieves strong performance with reliable effort prediction and interpretable cost estimation. The system can assist software project managers in making better planning decisions during the early stages of software development. This work demonstrates the effectiveness of combining machine learning techniques with traditional estimation methods to build a practical software effort estimation system suitable for academic and industrial environments.