Expand functionality with machine learning

[osanwe]

Hello, following up our previous discussion with @Min-Ho-Kim,

We have a tool for predicting bugs in software written in C/C++ with machine learning algorithms using metrics of source code. Our approach uses function-level and class-level metrics such as LoC, Halstead measures, Cyclomatic complexity, Measure of aggregation, etc. Decision trees, Logistic regression, SVM, Neural networks are used for defect prediction.

You already have a module for getting metrics [1], and we propose to expand it with our functionality.

If you are interested in it we can create a pull request.

Thank you.

[1] https://github.com/Samsung/Dexter/tree/master/project/dexter-metrics

cc: @arogozhnikov; @lucenticus

[Min-Ho-Kim]

Welcome to Dexter! And thank you for your interest.
After reviewing your pull-request, we will decide whether merge or not.

Plus, I've read your document that explains the idea of 'Software Defect Prediction".
That was very similar our vision of Dexter.
Thanks again.

[KarolAntczak]

Hi @osanwe!
Your proposition looks very interesting.
Do you have any paper that describes the approach? I would like to know how it was trained, what is accuracy of the model etc.

Regards,
Karol

[Min-Ho-Kim]

Hi @KarolAntczak,

I will send you the document that I received.
After pull-request, please review the codes.
I will give you and your co-workers reviewer permission,
so please let me know your co-workers IDs

@osanwe ,
Please do pull-request.
If you have other documents to share, please let us know.