Decision Model Based Reliability Prediction Framework

Abstract

Under every situation, neither the specific pattern model could be used, despite extensive data analysis attempting to expand reliability models of software. To make better use of current modeling techniques that are in usage such as combination and the model selection process, several other latest software reliability researches had been opted. Ineffective software reliability prediction is caused when incorrect model selection or weight allocation is trained frequently. This results in overrunning of the schedule. For combining various software reliability models on the keystone basis of multi-criteria decision trees, we postulate a methodical framework for prediction of reliability in this research paper. Based on experimental trends of multi-criteria sourced from multi reliability concepts, the method for model selection is suggested. For better and instantaneously allocation of weight per model, the decision tree with diminished defect edging ability deems the models with the predictive patterns preferred to be better. In this paper, investigation is done on the prospect of over- or under-prediction of the recognized models and the productive models in both predilected kind groups are weighted together. The proposed method exceeded current methods in terms of prediction accuracy rate, according to the results of analysis.