Web Based Application Maintenance Time Estimation Modeling by Bayesian SEM

Abstract

The objectives of this research were to find out the coefficient and other parameter estimation under unknown distribution and compare new model’s coefficient accuracy with [19]. This new coefficient and parameter were estimated with Bayesian analysis instead of Maximum likelihood. The model used in Bayesian analysis was start from the ML-model result from [19]. The new values were replaced into the former model then MMRE was detected from 30 completed software maintenance projects. The result of cross validation was about 40.32% while the ML-model was 42.55%.