Deep-SE is written in Python 2 and is organized in three modules, data, NEC, and classification.
This package is tested with Python 2.7.18. To run Deep-SE, you need to install the following packages first:
- Keras version 1.0.6 (
pip install keras==1.0.6) - Theano version 0.9.0 (
pip install theano==0.9.0) - Numpy (tested with version 1.16.6 ->
pip install numpy==1.16.6) - Pandas (tested with version 0.24.2 ->
pip install pandas==0.24.2) - SciKit-learn (tested with version 0.18.2 ->
pip install scikit-learn==0.18.2)
You can install all the required dependencies by running the following command at the current directory:
pip install -r requirements.txt
You also need to tell Keras to use Theano in its back-end. This can be done by setting "backend" attribute in keras settings json file to "theano". If you are on a UNIX like system (MacOS X or Linux), you can find keras.json file under /home/[your username]/.keras/. If there is no such directory or file, you can create one. For more help on this, click here: how to switch backend with keras from tensorflow to theano.
The dataset files are stored in ../datasets directory.
Run python data/run_script.py to split the dataset files to train-validation-test sets and tokenises the textual information to be used by Deep-SE in later steps.
For detailed information regarding how the data module works and how to configure it for different datasets, please refer to data/Readme.md.
If you wish to use pre-training, after running python data/run_script.py with setting the dataset to Pretrain_Dataset, run python NCE/exp_lstm2v.py with mode set to lstm2vec at line 3 of the script.
You also need to set the preferred dataset by changing line 80.
For more information on this module please refer to NCE/Readme.md.
Now you can run python classification/exp_script.py to perform issue story point estimation using Deep-SE.
For more information on this module please refer to classification/Readme.md.