A CAD System for Software Effort Estimation

The mentioned article investigates the use of approaches and techniques of machine learning especially the unsupervised learning ways in order to compute effort of software tools and application. Here, I have made use of mainly the historical dataset in collaboration with unsupervised learning techniques to predict the effort. First of all, to cluster the projects from the historical dataset then utilized the unsupervised category of machine learning approach as m-medoids in clustering with multiple similarity measures. Apart from this, we facilitate the results by using machine learning approaches to anticipate software effort by imputing missing values from past records. Furthermore, it is being experimented using ISBSG and CSBSG dataset which demonstrate that unsupervised learning as k-medoids clustering produced superior performance. In addition to this, it has been observed that kulzinsky coefficient proved to be best in estimating the similarities index of projects and imputation of missing imputation enhanced the efficiency of unsupervised approach of machine learning in software effort prediction.