An effective method for detecting duplicate crash reports using crash traces and hidden Markov models
K Neda Ebrahimi, Md. Shariful Islam, Abdelwahab Hamou‐Lhadj, Mohammad Hamdaqa · Computer Science and Software Engineering · 2016
When a software system crashes, crash information from user's machine is sent to the developers of the system for repair. For software systems with a large client base (such as Eclipse, Web browsers, etc.), the number of reports that are submitted every day can be quite high. Managing these reports is known to be a tedious and a time consuming task. Fortunately, not all crashes are caused by new faults. Studies have shown that around 30% of the reported crashes are duplicates of previously reported ones. Automatic detection of duplicate crash reports can then reduce the time and effort dealing with crash reports. In this paper, we introduce a novel method for detecting duplicate crash reports using crash traces and Hidden Markov Models. We show that our approach outperforms existing methods in detecting duplicate crash reports.