I am a Data Scientist need `jira data` to predict the `Severity` of the ticket

[ramesh_babu NA]

Hello All,

 I used to be a jira admin for one of a forutue 500  company, which has huge number of projects say 5000 and huge data say 4+ millions of tickets. ( I am not working there now .. ) 

During my experience, it was always an issue to nail the Severity  of the ticket / bug/ production bug. 

I am in the middle of changing my career to Data scientist . and I like to build a model which can predict the `Severity say 1 to 5` . I am looking for data from atlassian  jira so that once trained I can be easy  implemented for client who is using jira 

Thanks for reading my question, you may not have any lead  for now .. when you find can you please add your comments 

Cheers

Ramesh 

[Bill Sheboy]

Hi @ramesh_babu NA 

I am confused by your question as to why training based upon data is needed. In my experience Severity is an attribute based upon the impact of a defect (incident).  Often, questions in the incident report decide the severity, eliminating the need for support team levels to guess (or argue).  For example, see the image below.

Best regards,

Bill

[ramesh_babu NA]

@Bill Sheboy  thanks for your response .. Yes I agree with your Incident Response matrix. 

Step 1: User creates a ticket say for jira-1

step 2: User thinks there are no work around available and mark it as Severity:1

step 3: support team know the workaround and gives work around and brings down the Severity:2

setup 4 : team fixes this issue and ticket is closed after 2 days .

-- I like to have 100,000 tickets like this then .. pass it on `AI Machine learning` using NLB .. and create / train the model 

after training the model: 

  scenario  1 : user create a ticket I am not able to reach say `Amazon web site` 

  scenario 2:   user create a ticket , saying at the checkout page, after clicking pay , getting error page

scenario 3: I am not able access project 'Abc` 

scenario 4: I am not able access bastion host,  and marks Severity:1 thinking network is completely down . 

thinking of algorithm to predict ,  

scenario-1 as Severity:1

scenario-2 as Severity:2

scenario-3 as Severity:3, or 4.

scenario-4 as ML says it is Severity:3, or 4. 

In this case users are not involved , we can train the system everyday and it can accurately predicts the classification .. 

since I am new to Data Science , I think at this point, sentiment in the sentence(s) ML Machine Learning can be trained to predict .. without Front Line team .

[Bill Sheboy]

Thanks for that information, and I respectfully suggest this is an education, intake, and level 0 support (self-help info) opportunity.

Usually support level 1 (help desk/call center) does not allow customers to set the severity.  The intake process asks questions which lead to the severity determination, often programmatically from information provided by the level 2/3 team.

And, with guided information of self-help and FAQs (level 0), many lower "severity" incidents are eliminated as customer education issues.

There are circumstances where intake leads to an incorrect severity.  Those may be caused by the development team providing incorrect assessment criteria at application turn-over to help desk people.  Over time, the number of those issues are reduced.

[Gonchik Tsymzhitov]

Hi @ramesh_babu NA , 

How exactly  do you want to predict ? 

as you know a lot of interesting algorithms are exposed in ML things. 

• Linear Regression
• Logistic Regression
• Linear Discriminant Analysis
• Classification and Regression Trees
• Naive Bayes
• K-Nearest Neighbors (KNN)
• Learning Vector Quantization (LVQ)
• Support Vector Machines (SVM)
• Random Forest
• Boosting
• AdaBoost

[ramesh_babu NA]

thanks @Gonchik Tsymzhitov , 

 I am looking for tickets with Summary , Description, comments with `Severity` mapped accordingly . planning  to build all models what you have mentioned above + some neural networks

unless we build and train , we cannot depend on  any ML .

 thanks

Ramesh 

[Vivek Abushi]

Hi @ramesh_babu NA,
Just curious about the work you mentioned above. How did it go? Do you mind throwing light on the technical details and blockers if any.

Thanks
Vivek