Abstract

Estimation of software cost is crucial and essential in efficient project management. Modern projects exhibit high complexity and nonlinear relationships, and inaccurate estimations lead to budget overruns, delivery delays and inefficient resource utilization. This study suggests an Extreme learning machine-based cost estimation. To handle scale variation and reduce bias this method includes preprocessing historical data and applies single layer feed-forward ELM architecture consisting of several hidden parameters. Particle Swarm Optimization (PSO) and feature selection is used for improved performance, and it is validated through performance metrics like MMRE and RMSE. This model provides lower estimation error, and the model maintains relatively faster training compared to traditional models which results in faster, scalable prediction accuracy and strong generalization capability in software project planning. Also, the predicted effort is further applied to estimate project budget and create a phase-based SDLC schedule.

Authors

V. Venkataiah, Sarmista Rath, Sambari Koushik Kumar
CMR College of Engineering and Technology, India

Keywords

PSO, Hyperparameter Tuning, Extreme Learning Machine, Cost