Detecting User-Perceived Service Failure in Mobile Applications via Mining User Traces
Mobile applications often suffer from service failure issues nowadays for various kinds of reasons. Developers usually pay more attention to those issues that are perceived by users and compromise the user experience. However, to our best knowledge, there is no satisfactory approach to detecting whether users have actually encountered service failure issues. In this paper, we propose a novel approach to detecting user-perceived service failure issues. By leveraging the frontend user traces, we build an app page model, and design an unsupervised detection algorithm to detect whether a user has encountered service failure. Our insight behind the algorithm is that when service failure occurs on an app page, users will backtrack and revisit the certain page. Preliminary evaluation results show that our approach can achieve good detection performance on a skewed dataset.