Grey prediction based software stage-effort estimation

Abstract

The software stage-effort estimation can be used to dynamically adjust software project schedule, further to help make the project finished on budget. This paper presents a grey model Verhulst based method for stage-effort estimation during software development process, a bias correction technology was used to improve the estimation accuracy. The proposed method was evaluated with a large-scale industrial software engineering database. The results are very encouraging and indicate the method has considerable potential.

Data accumulation and software effort prediction

“…They concluded that the use of several incremental predictions related to iterative releases was more effective than a global, project-level prediction of effort. Finally, Wang et al [13] promote the use of a machine learning method utilizing small-sample grey models to retrospectively predict software effort per month over a range of largescale projects, concluding that such an approach could be useful for within-project adjustment of plans and resource management. While few in number, all of these studies point to the potential of within-project analysis that takes time into account.…”

Simulation Error Characteristics of Grey Model GM(1,1) under Translation Transformation

“…The model precision is measured by MARE in many researches [1,2,[5][6][7][8][9][10][11][12][13]. Though it is a most common used criterion in precision evaluation, it is sensitive to small item values and unfair to over-estimation [13].…”

“…The grey system theory uses the grey models to simulate the system's behavior and make predictions [1][2][3][4][8][9][10][11][12]. For reducing the complexity of the grey model building process, we can transform the original data sequence with translation and multiple transformations.…”

Integrate the GM(1,1) and Verhulst Models to Predict Software Stage Effort

“…Although prior research has failed to find clear laws by observation [29], [30], these works have shown that by using GM (1,1) or Verhulst grey models to make predictions using historical project data grouped by development methodology (or industry, etc. ), the mean prediction biases for each group are significantly different.…”

“…Using it, GV can fit the particular stage-effort sequences better and obtain more accurate results. In prior research [29], [30], a bias correction method has been used, but it tends to overfit. In this study, we make improvements based on the prior work, and propose a novel adjustment coefficient SAC as follows:…”