Smartphone Context Event Sequence Prediction with POERMH and TKE-Rules Algorithms

Smartphone applications run in complex environments and are sensitive to context events, i.e., changes to screen orientation, location, battery, etc. Context events and sequences make reliability, accuracy, and testing of mobile applications costly and more application development susceptible to errors. Modern smartphones have continued to increase the number of context event possibilities compared to earlier versions. Multiple context events may occur within quick intervals and complicate an application's behavior. While apps are continuously released and updated, future devices will likely have more features with more context events that will lead to even greater challenges. In this work, we implemented a model for prediction of common context event sequences by using two sequence rule mining algorithms: POERMH: Partially-Ordered Episode Rule Mining with Head Support, and TKE-Rules: Top-K Episode mining, along with a sequence prediction algorithm to increase smartphone application testing efficiency. Indeed, testing frequently encountered scenarios often relates to user perceived reliability of an app. The experiment was significantly successful at predicting next sequence in sequence of events for individual events as compared to overall performance using both POERMH and TKE-Rules in terms of Recall, Precision and F-1 score.