Combination of Predictions Obtained from Different Software Reliability Growth Models
In the development of techniques for software reliability measurement and prediction, many software reliability growth models have been proposed. Application of these models to real data sources has shown that there is commonly great disagreement in predictions, while none of them has been shown to be more trustworthy than others in terms of predictive quality in all applications. Recent work has largely overcome this problem through the development of specialized techniques which analyse the accuracy of predictions from reliability models. Such techniques allow the user to choose, for future predictions for a particular data source, those models which gave the best predictions in the past, for this data.
In this paper, various methods are used to get new predictions by combining the predictions obtained from different models. For each data set, the weights used in the combination of the models for each prediction of future data are based on the accuracy of the past predictions from the different models on this data. The resulting predictive quality of the combined predictions is investigated by application of these techniques to some real failure data. By using the combined prediction method it is demonstrated that improved predictions, or automatic selection of the “best” prediction system from all available prediction systems, can be achieved. An important benefit of the combination method presented in this paper is that there are no specific requirements on the nature of the initial prediction systems being combined.
Unable to display preview. Download preview PDF.