Software Crash Study

Abstract

With the development of personal computers, the user experience has become a vital part of every day work and life of the majority of people on the planet. Hardware components are usually preconfigured and most people tend not to tune them. However, the software environments change much more often because of the configuration by users, the upgrading by vendors and the attacks by hackers. All of those activities can be a factor in the stability of software. In this work, by analyzing a sample of 600,000 machine weeks and around 16,000 applications used on them, we try to derive the relationship between the software environment and the crashes of software. We mainly used association rule mining and analyzed our data on Spark. We also examined the predictability of crashes using the association rules and the difference of predictability between different versions of a same application.