基于改进的新陈代谢GM (1,1)模型的软件阶段成本预测 The Prediction of Software-Stage Effort Based on Improved Metabolic GM (1,1) Model

目前，关于软件成本预测的研究主要集中在对总成本的预测，对软件项目阶段成本的预测较少，然而软件行业对此有强烈的需求。为此，本文研究了使用灰色理论的GM (1,1)模型进行软件阶段成本的预测，并对GM (1,1)的新陈代谢模型进行了改进，动态选择模型初始条件，并提出了一种软件项目阶段成本的预测方法IGM。在三个不同数据集上的实验证明IGM方法优于传统新陈代谢GM (1,1)模型、GV方法和LR模型，显示出较大的潜力。 At present, the researches of software effort prediction mainly focus on the prediction of total effort, and the prediction of software project stage effort is less, but the software industry has a strong demand for it. So, this paper studies software-stage effort prediction by using the GM (1,1) model of grey theories, and improves the metabolic model of GM (1,1), selects the initialization dynamically, and proposes a prediction method IGM. Experiments on three different datasets demonstrate that IGM method is superior to traditional metabolic GM (1,1) model, GV method and LR model, and has greater potential.