DeepFL is a deep-learning-based fault localization technique. It implements two multi-layer perceptron variants (MLP and MLP2), one recurrent neural networks variant (BiRNN) and two tailored MLP variants by Tensorflow. The benchmark subject is from Defects4j, which is an open source repository, providing some buggy versions and corresponding fixed versions of different projects. Features of the dataset include different dimensions, e.g., spectrum-based, mutation-based, complexity-based (code metrics) and textual-similarity-based information.

• Python 2/3 with Pandas and Numpy
• Tensorflow

The dataset can be downloaded from an online Cloud Drive. There are seven .gz files, each of them represents one setting in our paper as follows:

• DeepFL.tar.gz: Dataset with all of four dimensions features above.

• CrossDeepFL.tar.gz: Dataset with all of four dimensions features in the cross-project scenario.

• CrossValidation.tar.gz: Dataset with all of four dimensions features for implementing 10-fold cross validation evaluation.

• DeepFL-Spectrum.tar.gz: Dataset with three dimensions features, i.e.,mutation-based, complexity-based and textual-similarity-based information.

• DeepFL-Mutation.tar.gz: Dataset with three dimensions features, i.e.,spectrum-based, complexity-based and textual-similarity-based information.

• DeepFL-Metrics.tar.gz: Dataset with three dimensions features, i.e.,spectrum-based, mutation-based and textual-similarity-based information.

• DeepFL-Textual.tar.gz: Dataset with three dimensions features, i.e.,spectrum-based, mutation-based and complexity-based information.

Please uncompress .gz files and put them into a user-created folder which can be passed as a parameter for the following command.

The command to run DeepFL for each version is as follows:

$ cd DeepFaultLocalization

$python main.py /absolute/path/to/ParentDirofDataset /absolute/path/to/Result $subject $version $model $tech $loss $epoch $dump_step

Each parameter can be explained as follows:

• /absolute/path/to/ParentDirofDataset: The absolute path of the parent directory including all datasets, for example, if the dataset is DeepFL, its directory can be /home/DeepLearningData/DeepFL ("/home/DeepLearningData/" is created by users, and "DeepFL" is put into it)
• /absolute/path/to/Result: The directory of the results.
• $subject: The subject name, which can be Time, Chart, Lang, Math, Mockito or Closure.
• $version: The version number of the subject. Note that, the maximum numbers of subjects above are 27, 26, 65, 106, 38, 133, respectively.
• $model: The implemented model name, which can be mlp, mlp2, birnn, mlp_dfl_1, mlp_dfl_2, representing multi-layer perceptron with one hidden layer, multi-layer perceptron with two hidden layers, Bidirectional recurrent neural network, two variants of tailored mlp, respectively.
• $tech: The different dimensions of features, corresponding to the name of dataset, can be DeepFL, DeepFL-Metrics, DeepFL-Mutation, DeepFL-Spectrum, DeepFL-Textual, CrossDeepFL.
• $loss: The name of loss function, which can be softmax, epairwise.
• $epoch: The number of training epochs.
• $dupm_step: The interval number of epoch in which the result will be stored into the result file. For example, if $dump_step = 10, the results in epochs 10, 20, 30... will be written into the files.

Please note that CrossValidation is slightly different with others since the dataset of all subjects has been mixed and then splitted into 10-fold. To easily use the command above, just set the parameter $subject as "10fold", $version as 1 to 10, and $tech as "CrossValidation". Also, please only use mlp_dfl_2 model and softmax loss function to run on CrossValidation according to the research question in our paper.

After running all subject versions, run the following command to calculate the five measurements Top-1, Top-3, Top-5, MFR, MAR:

python rank_parser.py /absolute/path/to/ParentDirofDataset /absolute/path/to/Result $tech $model $loss $epoch

Please note that due to the randomly initialized parameters, the results may be slightly different from our paper.