Uncertainty Measurement Method of Software Effort Estimation
Ye Yang · Jisuanji gongcheng · 2012
This paper proposes a method to assess the uncertainty of using COCOMO81 model.It uses the variance of model's output to assess the uncertainty of model's output.And uses both the data's variance and the model's output to assess the uncertainty of the distance between estimated effort and its real value.Based on the uncertainty,it proposes the prediction interval.It proposes the Mean Relative Variance(MRV) and shooting ration as the measurement to assess the uncertainty.In the experiment,re-Sampling method is employed to get N pairs of train set and test set.The average value of SR and MRV for different confidence level is calculated.After that,the scatter plot of SR and MRV is used to compare the different intervals that generated by different methods of assessing uncertainty.Experimental result shows that the measurement for uncertainty can generate narrower interval at the context of the same shooting ratio.