Maintenance costs can be substantial for organizations with very large and complex software systems. This paper describes research for reducing anomaly report turnaround time which, if successful, would contribute to reducing maintenance costs and at the same time maintaining a good customer perception. Specifically, we are addressing the problem of the manual, laborious, and inaccurate process of assigning anomaly reports to the correct design teams. In large organizations with complex systems this is particularly problematic because the receiver of the anomaly report from customer may not have detailed knowledge of the whole system. As a consequence, anomaly reports may be wrongly routed around in the organization causing delays and unnecessary work. We have developed and validated machine learning approach, based on stacked generalization, to automatically route anomaly reports to the correct design teams in the organization. A research prototype has been implemented and evaluated on roughly one year of real anomaly reports on a large and complex system at Ericsson AB. The prediction accuracy of the automation is approaching that of humans, indicating that the anomaly report handling time could be significantly reduced by using our approach.

Bug report assignment is an important part of software maintenance. In particular, incorrect assignments of bug reports to development teams can be very expensive in large software development projects. Several studies propose automating bug assignment techniques using machine learning in open source software contexts, but no study exists for large-scale proprietary projects in industry. The goal of this study is to evaluate automated bug assignment techniques that are based on machine learning classification. In particular, we study the state-of-the-art ensemble learner Stacked Generalization (SG) that combines several classifiers. We collect more than 50,000 bug reports from five development projects from two companies in different domains. We implement automated bug assignment and evaluate the performance in a set of controlled experiments. We show that SG scales to large scale industrial application and that it outperforms the use of individual classifiers for bug assignment, reaching prediction accuracies from 50 % to 89 % when large training sets are used. In addition, we show how old training data can decrease the prediction accuracy of bug assignment. We advice industry to use SG for bug assignment in proprietary contexts, using at least 2,000 bug reports for training. Finally, we highlight the importance of not solely relying on results from cross-validation when evaluating automated bug assignment.

Topic models, and more specifically the class of Latent Dirichlet Allocation (LDA), are widely used for probabilistic modeling of text. MCMC sampling from the posterior distribution is typically performed using a collapsed Gibbs sampler. We propose a parallel sparse partially collapsed Gibbs sampler and compare its speed and efficiency to state-of-the-art samplers for topic models on five well-known text corpora of differing sizes and properties. In particular, we propose and compare two different strategies for sampling the parameter block with latent topic indicators. The experiments show that the increase in statistical inefficiency from only partial collapsing is smaller than commonly assumed, and can be more than compensated by the speedup from parallelization and sparsity on larger corpora. We also prove that the partially collapsed samplers scale well with the size of the corpus. The proposed algorithm is fast, efficient, exact, and can be used in more modeling situations than the ordinary collapsed sampler.

We suggest a Bayesian approach to the problem of reducing bug turnaround time in large software development organizations. Our approach is to use classification to predict where bugs are located in components. This classification is a form of automatic fault localization (AFL) at the component level. The approach only relies on historical bug reports and does not require detailed analysis of source code or detailed test runs. Our approach addresses two problems identified in user studies of AFL tools. The first problem concerns the trust in which the user can put in the results of the tool. The second problem concerns understanding how the results were computed. The proposed model quantifies the uncertainty in its predictions and all estimated model parameters. Additionally, the output of the model explains why a result was suggested. We evaluate the approach on more than 50000 bugs.