An Ensemble Model for Software Development Cost Estimation

Software cost estimation is a critical activity that must be fulfilled in order to ensure the successful completion of the software project. Conventional methods of software cost estimation suffer from subjectivity and overhead. To overcome these limitations in the traditional approaches, Machine learning-based techniques have been emerged as an effective solution for the problem of software cost estimation. A stacking ensemble model has been proposed in this study as an effective approach for software development cost estimation. The proposed model is composed of diverse base learners and a meta-learner which is the Random Forest. The model has been trained using the ISBSG dataset which is a high-quality, heterogenous, and well-maintained dataset. The hyperparameters of the model have been tuned using the PSO algorithm. The experiments revealed that the stacking ensemble model outperforms the single contributing model and that various data preprocessing techniques can have a considerable influence on prediction accuracy.