[openstack-dev] [nova] bug triage experimentation

Nematollah Bidokhti Nematollah.Bidokhti at huawei.com
Wed Jul 26 01:54:11 UTC 2017


Following are the links from our team activities, please review and let me know how we can collaborate:

Python code is uploaded on Git repo and data set is available to download from dropbox.

We have used jupyter notebook interface to implement model.

Note: Download everything in same folder. 

Please follow the links:

Data Cleaning and Machine Learning Algorithm: https://github.com/ZainabSayyed/fault_genes/blob/master/OpenStack_n_gram_project.ipynb

Data Set: 
	Ask_os : https://www.dropbox.com/s/gowofd0nn7oa78j/ask_os.csv?dl=0
	Launchpad: https://www.dropbox.com/s/81k3g76s6y78pux/buglist.xlsx?dl=0
	Stackoverflow: https://www.dropbox.com/s/rxf469xd61s7ixm/report.xlsx?dl=0

Clustered Data Set:

	Ask_os: https://www.dropbox.com/s/u54rq7eo1xcgce4/df_ask.csv?dl=0

	StackoverFlow: https://www.dropbox.com/s/ul6gtfeykd328w6/df_st.csv?dl=0

	Launchpad: https://www.dropbox.com/s/iwoienwfd94u16y/df.csv?dl=0

Clustered Visualization: pyLDAViz 


-----Original Message-----
From: Sean Dague [mailto:sean at dague.net] 
Sent: 2017年7月21日 18:22
To: openstack-dev at lists.openstack.org
Subject: Re: [openstack-dev] [nova] bug triage experimentation

On 07/20/2017 06:20 PM, Nematollah Bidokhti wrote:
> Hi,
> I have missed the original email on this subject.
> We [Fault Genes WG] have been doing some machine learning analysis on Nova bugs/issues from 3 different sources (Launchpad, Stackoverflow, ask.openstack.org). We have been able to take all the issues and bring them down to 15 clusters.
> We have tried to find open source tools that can help us define the fault classifications, but have not been able to find any tool.
> Therefore, our team have come to the conclusion that we need the support of some Nova experts to help define the classifications. I would like to have some discussions with Sean and others that have an interest in this area and compare notes and see how we can collaborate.
> The goal of our WG is to apply the same technique to all key OpenStack projects.

Sure, would be happy to. All this went a little bit on hold as I was off on vacation and a conference, and am now trying to help getting freeze critical patches back in. But I'll probably start looking again more deeply after the feature freeze.

If you can provide me pointers to your machine learning work that clustered things, I'd happily take a look and see how that matches with domain experts. Thanks a bunch for also diving in here!


Sean Dague

OpenStack Development Mailing List (not for usage questions)
Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe

More information about the OpenStack-dev mailing list