Inferring certification metrics of package software using Bayesian belief network

Due to the rapidly increasing package software products, the quality certification has been required for software products. When certifying software products, one of the most important factors considered is the selection of the metrics. In this paper, specific package software types are represented as characteristic vectors having relationships with the specific metrics. The relationships are also described using probability. Once represented with the characteristic vectors, a specific package software product can distinguish itself from the other package software products. In order to utilize the past package software certification data, Bayesian belief network (BBN) is adopted. When using BBN, the dependency relationship network of the characteristic vectors and metrics should be constructed by first using the past package software certification data. The dependency relationship network is then used to infer the proper metrics for the certification of new package software products.