An exploration of techniques to Mine Patterns in Source Code

Details

Supervisors

Mens, Kim ; Nijssen, Siegfried ; Lienard, Julien

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique, à finalité spécialisée

Abstract

Every day, an increasing amount of source code is produced by students, scientists and companies, both proprietary and open source. To better understand programming habits and common errors, it is becoming increasingly essential to have tools capable of automatically analysing this source code. In a previous work, a pattern mining technique was developed, namely the exploration of ordered frequent subtrees called FREQTALS. However, the use of tree structures has a major weakness: each pattern found is very detailed, and we have many similar patterns found. In this thesis, we will explore another exploration technique, sequence mining. We will study how to represent the code for this technique and analyse the results to see if we can identify relevant and useful patterns in the source code.