UIChecker: An Automatic Detection Platform for Android GUI Errors
Meichen Ji · 2018
At present, Android automated GUI testing has been widely used in mobile application testing. Automated GUI test input generation technology and tools are hot topics for practitioners, but errors in some test screenshots generated by automated test input tools still need to be reviewed manually. In this paper, we creatively proposed an automatic detection platform for GUI errors, detecting the GUI errors of mobile related and image-related widget error classification model through machine learning, which detects the error of widgets. On all experimental App test sets, the accuracy of the text-related widget error classification model reached an average of 98.06%, and the accuracy of image-related widgets error classification model achieved an average of 95.44%, which greatly reduced the time cost of reviewing GUI errors manually. In addition, we analyze the relative positional relationship between the widgets, and use the Wilson score sorting algorithm to analyze the symbiosis and interdependence between the widgets, and finally generate the assertion tables, thus more complex GUI errors can be detected.