Skip to content

Latest commit

 

History

History
28 lines (22 loc) · 860 Bytes

README.md

File metadata and controls

28 lines (22 loc) · 860 Bytes

Introduction

prober is an automated software bug predictor using machine learning techniques. Specifically, proper trains prediction model based on open source datasets, and then predicts whether some code has a bug or not. The prediction is based on various code features, such as McCabe features, Halstead features, etc.

How it works

  • traning data is from open source dataset
  • training algorithm is based on tensorflow and keras
  • features extraction algorithm is specifc for each feature and each programming language. For java, the feature extraction algorithm can be found in java

How to use

traning

Install dependencies:

pip install -r requirements.txt

Start traning:

python -m prober.prober

Extract java features:

gradle build && gradle execute

LISENSE

MIT