Towards a Structural Clone Based Recommender System

Abstract

Structural clones cover all kinds of large-granularity repeated program structures such as similar methods, classes, directories, and their recurring combinations. We can use these structural clones to enable code completion by means of identification and recommendation of suitable candidates from a large code repository. By providing a user with recommendations based on Method Clone Structures (MCS) – a type of structural clones – mined from a large code repository, we are increasing our chances of recommending a set of methods that the developer is highly likely to use. The market basket analysis philosophy is implicitly manifested in our recommendation approach.

Publication
2016 IEEE 23rd International Conference on Software Analysis, Evolution, and Reengineering (SANER)