Use Case Points based software effort prediction using regression analysis

Software development effort prediction was an important stages in project planning. Poor prediction would lead to project failure, losing tenders and reduced profits. Several studies have improved Use Case Points as the effort prediction model using regression analysis. However, evaluation performance on the prediction models were biased and produce an asymmetric error distribution. Moreover, the dataset used were primarily from industrial, and less from universities. This study aims to investigate the performance of the regression model in terms of software development effort prediction based on Use Case Points using standardized accuracy (SA) and effect size (Δ) as the evaluation measurement. From the experiment results, regression model yielded 92%-0.64, 96%-1.86, and 69%-0.53 in term of SA and (Δ) over dataset DS1, DS3, and DS4, respectively. Experiment results shows that regression model yielded the best accuracy compared with the Karner model over three dataset. In the future, our results maybe used in development of effort prediction framework for calculating software project costs.