Data Collection Prioritization for System Quality Analysis

When assessing software quality the cost of collecting the data needed for analysis is often quite substantial. This paper proposes the use of Bayesian networks for assessing software qualities combined with an algorithm for how to prioritize which data to collect in order to minimize the cost of the assessment. This algorithm, the Diagnosis algorithm, is implemented in the Bayesian network tool 'GeNIe'. An example evaluation of service interoperability in the paper demonstrates that using the algorithm may reduce time spent on data collection significantly.