Machine learning methods with feature selection approach to estimate software services development effort

Estimate of the effort required for software services development has been a most important topic in the field of service in recent years. Exact estimate of effort is a key factor for project's successful management and control. Over and underestimation waste system resources endanger the position of the related company. The development effort estimation is done with the help of expert judgement, algorithmic and machine learning methods. Recently, several methods of machine learning have been used to estimation software services effort and look much better than the other two groups. This paper presents an experimental evaluation of the effectiveness of these methods with feature selection approach and done a thorough comparison of their accuracy. Evaluation and comparison have been made onto two famous datasets NASA and ISBSG and results are well demonstrated position of each one of these methods.