SOFTWARE EFFORT ESTIMATION USING MACHINE LEARNING

Abstract

Software cost and effort estimation is a critical task in software engineering. Research in this field continually evolves with new techniques, requiring periodic comparative analyses. Accurate software cost estimation is vital for project success, as it helps identify the challenges and risks involved in development. The wide variety of machine learning (ML) and non-ML techniques has led to comparisons and the integration of these methods, making it essential to determine the most effective estimation techniques to enhance the project development process.

The review follows a three-stage approach: planning (Tollgate approach), conducting (Likert-type scale), and reporting results from five well-known digital libraries. Among the 52 selected articles, the artificial neural network (ANN) model and the Constructive Cost Model (COCOMO) have been identified as the most favored techniques. The mean magnitude of relative error (MMRE) is the preferred accuracy metric, with software engineering and project management being the most relevant fields. The PROMISE repository has emerged as the most widely accessed database.