A Distribution based Approach of Outlier Removal for Software Effort Data
Amit Sinhal, Bhupendra Verma · International Journal of Computer Applications · 2013
There are unavoidably a few outliers in the software measurement data.When software effort estimation models are made using the data samples with outliers, these models reduce the effort estimation precision for future planning.Therefore, this work investigated the influence of outlier upon the accuracy of prediction and proposed a distribution based outlier elimination method for effort estimation.The proposed work shows that the applied outlier elimination method improves the estimation accuracy of the software effort estimation process.In contrast, the effects of outlier elimination on the accuracy of effort estimation may differ depending on the characteristics of the data set, the effort estimation method.