back to notes

Machine Learning in IT

Machine Learning in IT

For the next phase we will be brainstorming data, problems and technologies.
Data
release calendar
ClearQuest
anthill
jenkins
server logs
splunk
clear quest
smokestack
BEV
social media
sonar qube
git
echo
general nelson
performance

Splunk:
access log
Problems
how many sev 1,2,3 in current build?
understand defect density to release calendar to flag release risk (clearquest source)
productivity improvement - automate
ML to help with code quality - automate test generation
track github review/reject trends for threshold alert, defect locations, code author, return suggestions that particular code needs review or other insights
code review - which PRs made it through, what changes - predetermine failure - robot code reviewer
team burnout detector

How can I predict that a platform - accounts / payments / servicing will be stable in production next week ?
identify code defect source for problem resolution and internal billing
classify defect as code vs reqs - ingest BRDs and compare to defect
monitor keywords in defects for urgency
analyze logs to determine what is a potential problem
code analysis - UTRM, security, code smells
automatic log grooming - splunk logs too big to be useful
Correlate customer feedback (social media, support calls etc) with build/GIT timestamp info to identify likely issue source
track system availability, correlate to time of day, etc. ML to understand common issues vs new pattern
monitor transaction volume to identify unusual behavior and alert potential problems
predict whether platform stable in a future timeframe
PTO/training predictor
customer sentiment analysis based on feature release
monitor email for priority - noise filter and highlights
Technologies



Solutions
Interview Questions
What data questions do you have to ask regularly?
What information do you think is in our data that we aren't using?
What are your biggest information pain points?


last updated september 2019