preke / apsec_2018

APSEC 2018

Codes and data for the paper Detecting Duplicate Bug Reports with Convolutional Neural Network in APSEC 2018

Requirement:

• Python3.6
• Anaconda(numpy, pandas, sklearn)
• PyTorch 0.4.0
• torchtext
• gensim
• cuda 8.0

Basic usage:

Before run codes, set parameters(paths) in each .py file in codes/

Traditional CNN

• Train model: python main.py
• Evaluate existed model:python main.py -snapshot *.pt (*.pt is the existed model)

DBR-CNN

• Generate DBR-CNN result: get into codes/[data_set]/[data_set] means the specific dataset you are using, like (spark, hadoop, hdfs, mapreduce)
• python cb.py

Use specific word vectors:

In main.py:

1. set use_global_w2v = False
2. set wordvec_save to specific .save file

Change CNN parameters:

In main.py: change variables straightly in parser

---

Pretrained word vectors could be download from https://pan.baidu.com/s/18R_lZhlOdp-kgDlbrBq7iA and unzip it in codes/