Font Size: a A A

A Learning Based Fault Localization Approach

Posted on:2019-05-16Degree:MasterType:Thesis
Country:ChinaCandidate:SYED RIZWANSYDFull Text:PDF
GTID:2428330566997330Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Localizing the faults is one of the major tasks in software debugging.Manually developers spend most of their power and time for locating the faults correctly.Hence to assistance the developers,automatic fault localization techniques are providing the role like the main bridge.Generally,in the area of fault localization,numerous techniques have been presented which inputs the test suits and outputs the list of faulty entities of a program which are ranked.In all of these techniques,likely program invariant is one of the dominant ways to discover and analyze the software bugs because by utilizing this,those characteristics can be captured to analyze which disrupt the goal of a developer on which they committed to.To improve the work in this direction,we have studied the fault localization approach based on invariants,and presented a system called SILearning,which localize suspicious methods in programs by learning from some existing fixed bugs.It combines a machine learning approach known as“learning to rank”,program invariant analysis,and spectra-based fault localization.Invariant difference and suspiciousness values which are examined on both test cases and code coverage analysis as features for ranking the faulty methods in the light of their probability of being a source of failure.To accomplish this work,we divide the SILearning into five parts:cluster of faulty methods and subset selection of test cases,likely and dynamic invariant detection,subset selection of invariant,extraction of feature and model learning and method ranking.After accomplishing all these steps,SILearning outputs a list of faulty methods which are ranked by suspiciousness.We have practically analyzed the SILearning on the dataset of real faults which is extracted from the database“Defects4J”and compare the performance of SILearning with the previous state of the art.We discover that SILearning performs better when combine features are utilized and can successfully localize the faulty methods on average for“76.1”,“90.4”,“108.2”,“123”and“143.5”at the top 1,2,3,4 and 5.We also analyzed that SILearning outperforms of other techniques(ER1~a,ER1~b,ER5~a,ER5~b,ER5~c,GP2,GP3,GP13,GP19 and Savant).
Keywords/Search Tags:Fault Localization, Learning to Rank, Subset Selection of Invariant
PDF Full Text Request
Related items