Convolutional Neural Network for Classification of Source Codes

Abstract

A method to classify source code based on convolutional neural networks is presented. The goal of the neural networks is to predict the type of algorithm that is used in the corresponding source code so that the result obtained can be used for different kinds of assistance and assessment for programming education. In the proposed method, source code is converted into a sequence that represents the structure of the code without any keywords, such as variable names or function names. In present paper, models and implementation of the proposed method are presented. An experiment considering several algorithm types is also conducted. For evaluation of the proposed method, source code accumulated in an online judge system is used. The results of the experiment demonstrate that the proposed method can predict the algorithm used in the given source code to a high degree of accuracy.