Empirical Assessment of Machine Learning Models for Effort Estimation of Web-based Applications

Effort estimation plays a pivotal role for planning the development of any software applications. When there is a need of developing web-based applications, the need of proper estimation of its effort is considered to be very essential. Web-based software projects are different than conventional projects, and hence the task of estimation is a complex one. It is observed that the literature does not provide a guidance to the analysts to use a particular model for effort estimation of this type of applications. A number of models like IFPUG Function Point Model, NESMA, MARK-II, etc. are being considered for effort estimation of web-based applications. The efficiency of these models can be improved by employing certain intelligent techniques on them. This study considers the application of various machine learning techniques such as K-Nearest Neighbor (KNN), Constrained Topological Mapping (CTM), Multivariate Adaptive Regression Splines (MARS) and Classification and Regression Tree (CART) for effort estimation of web-based applications using IFPUG Function Point approach. The ISBSG dataset, Release 12 has been considered in this study for obtaining the IFPUG Function Point data. The effectiveness of these machine learning techniques-based effort estimation models is assessed with the help of certain metrics, in order to evaluate performance of desired web-based applications.