Analysis of Mousavian Technique’s Scalability for Software Bug Localization

• Main Author: Umar Fariz Tumbuan, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39262
• Institution: Institut Teknologi Bandung

Bug localization technique is a concept that was developed to improve the efficiency of debugging process in software development. One type of bug localization technique with a great result are techniques utilizing the graph mining algorithms. Unfortunately, graph mining algorithms do not scale towards the size of graphs. Mousavian et al. (2011) developed a bug localization technique utilizing a new graph analyzing technique which is more scalable compared to graph mining based techniques. However, mousavian technique has not been tested using complex graphs as inputs. In addition, the algorithm to generate suspicious paths in the fourth phase of mousavian technique was also not yet explained in detail. This experiment aims to find mousavian technique’s process time and memory requirements growth pattern to determine its scalability, as well as to compare its process time with FSG, gSpan, and CloseGraph graph mining algorithms when using synthetic test data. Analysis result concludes that breadth first search (BFS) algorithm and depth first search (DFS) algorithm can be utilized to implement the suspicious path generation algorithm. DFS algorithm has an advantage over BFS algorithm due to its efficient memory requirement growth. In addition, the choice between DFS and BFS must consider the posibility of reducing mousavian techniques‘ scalability. One approach to prevent this is by limiting the travel depth of DFS or BFS. The test data are complete graphs with their predicates ranging from 100-3000. A graph data is represented as an adjacency matrix with its edge weights generated randomly. Analysis and test result shows mousavian technique’s process time and memory requirement both have polynomial growth pattern. Based on that results, mousavian technique can be categorized as a scalable bug localization technique.