Automated library recommendation

Many third party libraries are available to be downloaded and used. Using such libraries can reduce development time and make the developed software more reliable. However, developers are often unaware of suitable libraries to be used for their projects and thus they miss out on these benefits. To h...

Full description

Saved in:
Bibliographic Details
Main Authors: Thung, Ferdian, Lo, David, Lawall, Julia
Format: Conference Proceeding
Language:English
Subjects:
Association rules
Collaboration
Feature extraction
Generators
Itemsets
Libraries
Citations: Items that cite this one
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Many third party libraries are available to be downloaded and used. Using such libraries can reduce development time and make the developed software more reliable. However, developers are often unaware of suitable libraries to be used for their projects and thus they miss out on these benefits. To help developers better take advantage of the available libraries, we propose a new technique that automatically recommends libraries to developers. Our technique takes as input the set of libraries that an application currently uses, and recommends other libraries that are likely to be relevant. We follow a hybrid approach that combines association rule mining and collaborative filtering. The association rule mining component recommends libraries based on a set of library usage patterns. The collaborative filtering component recommends libraries based on those that are used by other similar projects. We investigate the effectiveness of our hybrid approach on 500 software projects that use many third-party libraries. Our experiments show that our approach can recommend libraries with recall rate@5 of 0.852 and recall rate@10 of 0.894.
ISSN:1095-1350
2375-5369
DOI:10.1109/WCRE.2013.6671293