关键词: 软件测试, 改进Gompertz模型, 缺陷, 预测

Abstract: Software testing prediction based on Gompertz model is poor fitness，this paper presents a novel model，in which the original Gompertz equation is expanded with the Taylor series by the Gauss-Newton method.After the initial value of parameters and sampling are chosen，the least squares method is used to calibrate parameters iteratively to estimate the parameters more accurately.The computing result with the actual data shows that this model improves the accuracy rate of defection detection，and that the absolute error rate in this model is less than in the original model.The computing result also shows that the fitness with the actual data in this model is better than in the original model.

Key words: software testing, improved Gompertz model, bug, forecasting