Loading...

Detecting Similar Android Apps Based on Users’ Reviews and the Source Code

Abedini, Yasaman | 2017

582 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 50211 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Heydarnoori, Abbas
  7. Abstract:
  8. Increasing popularity and use of smartphones has led to rapid growth in number of mobile applications. With the increasing number of available apps in app stores, some of them may have many similarities to each other in functional aspects. Detecting similar apps among a large number of different applications is an important and difficult problem which is useful in a variety of applications, such as app recommendation, comparing competitor apps, app search engines, etc. In order to investigate the functional similarities between applications, high-level and low-level characteristics can be used. With using the high-level characteristics, app functionality features can be extracted from valuable information available in the app stores, such as app names, descriptions, reviews, etc. And also app similarities can be detected by using source code features as low-level characteristics. Previous researches in this area have been focused on using one of these two groups of characteristics. However, using one of them alone may have some disadvantages such as the effect of not having enough descriptions or reviews, the large volume of extracted features, computational load for additional comparisons, etc. In this thesis, we first cluster android apps with the help of features extracted from the names, descriptions, and reviews (crawled from Google Play), then rank similar apps in same cluster by using features extracted from the source code along with other features. The evaluation of this approach indicated that 84% of features extracted from Google Play metadata was related to apps functionalities. In addition, this approach identified similar apps (top-5 results) with 80% precision, which has improved Google Play results on same dataset by almost 15%
  9. Keywords:
  10. Opinion Mining ; Android Applications ; Application Stores ; Similar Applications ; Application Features

 Digital Object List

 Bookmark

No TOC