Data mining to proactively capture trends in user problems

[rmackay9]

Some hardware or software problems are so rare that they can only be spotted by noticing trends in failures. The dev team (Randy, Tridge, etc) would like to have the Auto Log Analyser run on all log files that are submitted and the results added to a results database.

Users should be able to opt-in or opt-out of participation of course.

Some extra information to help correlate problems with other factors would be helpful. I.e. board manufacturer or location, software version, frame type, etc.

[AndyLeeRobinson]

Suggest using gzip compression too if not used already.

[mrpollo]

Thanks @rmackay9 Auto Analyze instead of on-demand should be the norm, and yes we should be able to expose some API to query our dataset of results of the analyzer. 👍

[geeksville]

we have a rest api to get the analysis results, but analysis is only done
lazily now. having it done automatically makes sense. (and yep - we've
always been using gzip for the streams from client to server)
On Tue Dec 23 2014 at 11:25:12 AM Ramon Roche [email protected]
wrote:

Thanks @rmackay9 https://github.com/rmackay9 Auto Analyze instead of
on-demand should be the norm, and yes we should be able to expose some API
to query our dataset of results of the analyzer. [image: 👍]

—
Reply to this email directly or view it on GitHub
#141 (comment)
.

[mrpollo]

i meant from a public data set, not per mission, /public or something similar, instead of /mission which would get all records then you would need to get analysis of each