PERFORMANCE ANALYSIS OF PREDICATED BUG SIGNATURE METHOD FOR BUG LOCALIZATION OF PROGRAMS

• Main Author: Ihza Parama, Rizki
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39385
• Institution: Institut Teknologi Bandung

This undergraduate thesis has a purpose to analyze performance and quality of bug localization with predicated bug signatures. Predicated bug signatures is one of the methods that can be used to do bug localization and to find bug signatures in the form of predicates set. Analysis is done with a study of the algorithm of predicated bug signatures method and also with an experiment. The performance of bug localization will be measured after experiment is done using different configuration of related parameters. By using many different configuration of related parameters, the relation of performance to each parameter can be evaluated. The algorithm of bug localization with predicated bug signatures consist of two steps which are generator mining phase and generator classification phase. Generator mining phase has recursive property because a search is done to find generator from combinations of predicate. Generator classification into equivalence class has traversal property relative to the generator set. Complexity analysis was also done to find the theoretical time and memory complexity of the algorithm. Time complexity is in order of factorial complexity relative to the size limit of bug signatures because of the recursion property. Memory complexity will be increased proportional to the depth of the recursion, number of predicates, and also the number of execution test case used. The experiment was done using a prototype of implementation of predicated bug signatures. The scheme and design of the experiment was to execute bug localization with many different test cases. Experiment was executed with different configuration for each of related parameters. For each parameter, the relation of that particular parameter to performance of the bug localization algorithm will be recorded. After all the data was recorded, the next step to do is to analyze the result of the experiment with statistical method such as calculating the mean and median. In order to increase the statistical explanation of the data, regression will be used to describe the shape of the data and also relation between them. The conclusion of this undergraduate thesis is the accuracy and the performance of bug localization with predicated bug signatures is affected by the parameters of the algorithm. The accuracy of this method can be increased as needed by increasing the number of execution test case for each data and also by reducing the number of bug signatures requested. The accuracy of bug localization is not affected by the number of predicate. The time performance of bug localization is significantly worsen by the size limit of generator because of the recursive property. The memory performance of bug localization is significantly worsen by the number of execution test case with the increase of memory used is in polynomian order.