Integrated Prediction Model for Software Effort of Diversifying Projects Using Machine Learning

Among the most significant responsibilities for those involved in software project management is software effort estimation. The fact that software development is constantly evolving makes it very hard to forecast effort. In the past, academics have estimated effort and duration for one type of methodology for software development. Various size matrices are used for software project estimation. To estimate the lines of code, use cases, objects, and story point for various development methodologies, algorithmic models are used. This work develops a hybrid software-estimating model for projects that are both object-oriented and traditional. There is the primary input is the size of the software, expressed as the lines of code and use case points. The combined Software Effort Prediction Model is created using linear regression analysis. This model is developed using the size of the project and the selected parameters identified with the correlation coefficient. The proposed work has been analyzed using the average magnitude of relative error. This study is further evaluated using existing methods for estimating software costs, and it is discovered that these methods—case-based reasoning, linear regression, radial basis function neural networks, ensemble modeling, and fuzzy analytic hierarchy process — the combined models have the least amount of inaccuracy. Linear regression is used to forecast the effort for procedural and object-oriented projects, and the findings are compared to those from other models to ensure the work is valid. The work obtains the highest accuracy for accurately estimating software projects.