Source Code Recommendation with Sequence Learning of Code Functions.

For finding desired source codes and articles using existing source code search engines, it is necessary to compare them in the different results to examine which source codes are the best practice for developing the target software. In this paper, we propose a method of source code recommendation based on the prediction of code function. The feature of the proposed method is that by interpreting the context of the processing procedure of the source code, it recommends source code which complements or follow the processing procedures. In the experiment using actual source codes, we evaluate the feasibility of the proposed method.