Advanced effort estimation in software projects using ensemble-based fusion techniques

• This research aims to propose a methodology to enhance effort estimation, ensuring that initial work estimation remains precise and reliable. In such cases, a fusion approach using an ensemble of machine learning techniques is proposed. The stacking concept is used to facilitate this ensemble/hybrid approach, allowing the learner model to learn from combined predictions. • The proposed methodology uses linear regression and random forest machine learning algorithms. The results are cross-checked against well-known datasets to demonstrate their estimation accuracy. • A machine learning method is used to forecast the project’s effort, as some parameters required for work estimation may not be determinable at the beginning of the process. The model is trained using parametric training with fixed-size parameters, with the use case repository data serving as the training set and the China dataset as the test set. • The proposed work uses an ensemble model to compute the fusion approach’s result, which is then inputted into a linear regression model. All regression models are then run individually to calculate the effort. The software development process heavily relies on effort estimation, which is crucial for determining the development approach and methodology. This research aims to propose a methodology to enhance effort estimation, ensuring that initial work estimation remains precise and reliable. In such cases, a fusion approach using an ensemble of Machine Learning (ML) techniques is proposed. The proposed methodology uses linear regression and random forest ML algorithms. The results are cross-checked against well-known datasets to demonstrate their estimation accuracy. A ML method is used to forecast the project’s effort, as some parameters required for work estimation may not be determinable at the beginning of the process. The model is trained using parametric training with fixed-size parameters, with the use case repository data serving as the training set and the China dataset as the test set. The proposed work uses an ensemble model to compute the fusion approach’s result, which is then inputted into a linear regression model. All regression models are then run individually to calculate the effort, and the efficiency of the proposed model is 94% which is a great improvement as compared to the individual model’s performance which is 74% only.