A fog computing based concept drift adaptive process mining framework for mobile APPs

Mobile applications are widely used to provide users convenient and friendly service experiences. Meanwhile, service logs generated by mobile applications are analyzed to obtain user behavior patterns for monitoring and optimizing mobile application performances. However, due to the frequent updates in mobile application, situations of concept drifts often occur in service log streams, which lead to challenges in mobile process mining. In this paper, a novel framework is proposed to solve the above problems by combining fog-computing-based concept drift detecting with cloud-computing-based process mining. Firstly, incomplete log data are preprocessed using fog-computing technologies to provide more accurate log contexts and lower overhead. Then, concept drift detecting methods are used in cloud computing layer to deal with the transfer of mobile applications from one version to another. Finally, experimental results demonstrate that our framework can deduce missed case identifiers for logs when concept drifts happen.