Automatic Categorization with Deep Neural Network for Open-Source Java Projects
Anh Tuan Nguyen, Tien N. Nguyen · 2017
In this work, we introduce a Deep Neural Network model for automated software categorization. The model is ableto form high-level concepts from low-level code tokens and to distinguish important features such as API calls and identifiers in order to support software categorization. Our empirical evaluation shows that DNN outperformed other machine learning approaches with 15.9-36.4% higher accuracy in software categorization. We plan to expand further our studies to explore more features and variations of DNN, with different configurations and data sets.