|Title:||User feedback assisted mobile apps analysis|
|Advisors:||Luo, Xiapu (COMP)|
|Subject:||Hong Kong Polytechnic University -- Dissertations|
Application software -- Development
|Department:||Department of Computing|
|Pages:||ix, 85 pages : color illustrations|
|Abstract:||With the rapid development of Android mobile devices, millions of mobile applications have been produced by developers. Software maintenance for mobile apps becomes an important and challenging task. User reviews and bug reports are two main forms of user feedback. Developers usually extract useful information from them to maintain and update their mobile apps. In mobile app maintenance, user reviews play a more important role than bug reports because user reviews can feed back information more immediately than bug reports. Unfortunately, reading reviews is tedious and time-consuming. For example, a famous mail app k-9 mail has more than 86 thousand reviews in Google Play Store. Thus, it is hard to read them manually. As far as we know, CHANGEADVISOR is the first work to utilize user reviews to assist developers to find which part of code should be improved. However, only using reviews to locate buggy code can lead to high false positives since reviews only contain little useful information. To handle this problem, we propose a novel approach which utilizes bug reports as a bridge to link user reviews and source code of apps. Bug reports on Github can provide more useful information like code examples, stack traces and patches which can improve the accuracy of bug localization. First, we use bug reports to build a topic model to verify whether a cluster of reviews describes real faults that are also reported by the bug reports. Second, we propose a weight-iteration cosine similarity metric to compute the similarity between bug reports linked to the cluster of user reviews and buggy classes. As a result, we get a ranked list of buggy files for the review cluster. The results of the experiments carried out on 31,399 reviews and 2,279 bug reports of 10 open source mobile apps show a high accuracy in buggy class localization. In addition, the comparison result of performance on faulty code localization shows that our approach is more accurate than the state of the art approach CHANGEADVISOR by up to 14.08% of precision, 19.39% of recall, and 21.6% of F-Measure.|
|Rights:||All rights reserved|
Files in This Item:
|991021953150803411.pdf||For All Users (off-campus access for PolyU Staff & Students only)||3.33 MB||Adobe PDF||View/Open|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: