Software Development Effort Estimation Using Machine Learning Techniques: Multi-linear Regression versus Random Forest

The software development industry has lately become quite intricate, at a global level. As the tools and technologies used keep changing, so does the approach of developing a software. Thus, software effort estimation plays a critical role in doing so. This arises a challenge of accurately estimating the software development effort, and then proceeding with the plan of development. The history shows various algorithmic cost estimation models like Boehm's COCOMO model, Putnam's SLIM, Multiple Regression, Statistical models, and many non-algorithmic soft computing models]. Despite multiple techniques, achieving a higher accuracy of effort estimation has always been challenging. This paper is concerned with a comparison between two algorithmic regression models, one using Multiple Regression, and another model using Random Forest Regression, to predict the estimation of software development effort. It is observed that Random Forest Regression is successfully able to model the complex, by closely matching the effort estimated in the dataset, providing a better accuracy.