摘要：Splitting identifiers is a task that has been addressed in the past few years in order to contribute toward improving the Feature Location task. Feature Location aims at determining the exact position of a specific feature within a source code. Several research studies have addressed the process of splitting multi-word identifiers. However, one of the endure gaps that still face the use of machine learning lies on using probabilistic algorithms which may seem insufficient compared to other sophisticated algorithms such as the Backpropagation Neural Network (BPNN). Therefore, this paper proposes a BPNN for the splitting identifiers task. A benchmark of source code dataset has been used in the experiments. In addition, different objective functions have been used including Tanh, Sigmoid and Softmax. Results showed that Softmax has outperformed the other objective funciton by achieving a 71.4% of f-measure. This results implies the usefulness of BPNN in terms of handling character-based problems.
Loading...
