ABSTRACT Software effort estimation (SEE) is widely recognized as one of the most challenging tasks in software project management. Researchers have tried various methods, including machine learning (ML), to address the SEE issue in the past. The ML domain is currently investigating the integration of individual models (ensemble models), resulting in models with superior performance. Bagging, boosting, and stacking are a few ensemble methods extensively investigated in the literature for enhanced SEE. The efficacy of an ensemble model primarily hinges upon the accurate optimization of hyperparameters for each base learner and the allocation of suitable weights to them. This research aims to develop a self‐adjustable ensemble method for improved SEE. The proposed method combines the problems of hyperparameter tuning and weight allocation into a single step and uses optimization methods to find solutions to both problems. Additionally, the proposed approach deals with the issue of data heterogeneity by selecting a distinct training subset for each testing project based on their similarity. The proposed approach employs the particle swarm optimization (PSO) algorithm for hyperparameter tuning, sequential least squares programming (SLSQP) for weight optimization, and K‐nearest neighbor (KNN) method for identifying the locality for each testing project. An investigation was conducted to evaluate the proposed model through an experimental analysis. The investigation utilized a range of datasets from the data repositories managed by PROMISE and ISBSG, which are well‐known SEE data sources. The results suggest that the performance of the proposed approach was better than that of the individual models. Furthermore, we compared the proposed strategy and other commonly utilized ensembling methods documented in the literature. Based on the results from our research, it can be concluded that the suggested methodology exhibits enhanced predictive capabilities compared to alternative ensemble techniques.
Shukla et al. (Wed,) studied this question.