Completion of Parallel app Software User Operation Sequences Based on Temporal Context

Abstract

While mobile application (app) software is becoming increasingly important in people's daily lives, researchers have the limitation of understanding the details of user operations inside the app. With the update of the Android system and application user interface, relying on manually defined user operation event templates or modifying the app source code can no longer meet the needs of fine-grained user operation analysis in multiparallel applications. In this article, a novel method is proposed for effectively analyzing user operations in parallel apps based on the temporal context of user operation sequences. The authors provide a general framework in the Android system to parse out fine-grained user operations. In addition, the authors build a deep learning model with LSTM-TextCNN to complete user operations in parallel app from global temporal context and app temporal context. The authors collected 240k operations of 12 users over a month. Comparative experiments with a baseline show that the proposed method can efficiently and accurately analyze parallel app user operations.