Software Development Effort Estimation using Machine Learning Techniques

Abstract

Software Development Effort Estimation is of great significance in the software industry as it plays a crucial role in ensuring the successful delivery of software solutions. Accurate effort estimation is essential for effectively managing the five dimensions (5D s) of the Software Development Life Cycle (SDLC): Demand, Development, Direction, Deployment, and Designated cost of the software. This involves forecasting the resources that are required for software project development, typically measured in person-month or person-hours. newlineSoftware Development Effort Estimation holds importance in various software development endeavors, including industrial software systems and digital transformation initiatives. Digital transformation involves integrating digital technology into different aspects of a company or organization to improve operations, procedures, customer experiences, and overall performance. On the other hand, industrial software systems refer to specialized software packages designed for use in industrial and manufacturing processes. newlineThe main objective of the presented thesis is to explore the application and stability of machine learning models to develop a robust and precise effort estimation model. newlineThe presented thesis aims to investigate the selection techniques on accuracy by identifying the most important and relevant features in the dataset while eliminating redundant and irrelevant ones. The presented work compared the performance of various machine learning models with different feature selection techniques. newlineThe thesis also proposes an Omni-Ensemble Selection (OSS) approach, which combines static ensemble selection along with a genetic algorithm (SES-GA) to extract the best suitable models for estimation of effort from the pool of models, and dynamic ensemble selection (DES) which later generates the final evaluated effort. The thesis also discusses a multi-step dynamic ensemble selection (MS-DES) approach to assist analysts and experts in their decision-making process.