Machine Learning Based User Interface Testing to Predict Defects Leading to Logical Errors

Traditional test suites for GUI applications do not cover all the functionalities of the graphical interfaces. Testing all the user interfaces in the GUI applications is a time consuming process. Use of machine learning to automate GUI testing process has increased the quality and the productivity of the software. This approach extracts the GUI controls and their properties using the GUIWT tool. Supervised J48 decision tree algorithm is applied for classifying the statements in to defect or no-defect category. The proposed work focuses on finding the logical defects in graphical user interfaces. Training data set is constructed from the various GUI applications. The defect prediction discover the logical errors in visual applications. The prediction model is empirically evaluated using various defect prediction metrics. The results show that prediction performance of J48 decision tree is better than the other defect detection techniques. This prediction model will help testers to produce defect free software.