All Apps and Add-ons

Can I "teach" the machine learning tool kit projects in our company converge through time, based on project name and the field?

matansocher
Contributor

Hi

I want to "teach" the machine learning tool kit how our projects in our company converges through time based on project name and the field 'SLOC'.
How can I tell the algorithm to see the different projects in our company and forecast the current project?
I can build a table that contains 3 fields: project name, date, sloc
Example:
project 1, 01.01.18, 22
project 1, 01.02.18, 40
project 1, 01.03.18, 2
project 2, 01.01.18, 400
project 2, 01.02.18, 88
project 2, 01.03.18, 16

Is it possible to "feed" the algorithm to know how our projects are going, so it will be able to forecast the next projects based on the previous project's data?

Thanks

0 Karma

DalJeanis
SplunkTrust
SplunkTrust

The answer to that is dependent on data and reality.

1) Do the projects have enough similarity in real life that the processes should track similarly?

2) Do you have enough data that machine learning is the correct methodology?

I suspect, from your questions, that the answer to the first question will be "maybe", and to the second question will be "No". This sounds like a case of wanting to try out ML, and looking for a use case.

Machine learning should be used when the data are both complex and large. This seems like you could use other techniques much easier.

But, before you even ask THAT question, you need to understand something. ML can't accomplish anything that a smart enough human can't accomplish, given time. So, before you try to throw a bunch of data at a machine instance, think the problem through from the point of if YOU had to figure it out. What would you need to know and want to know? Then ask for that data, or the closest you can get.


Knowing nothing else about your use case, here's what I'd do - first, I'd think about what the SLOC number means, in the real world. Source Lines of Code, the standard meaning, would be insufficient for predicting anything. It would be like: here's our record of dollars spent building each house. please predict when the next house will be complete. Spent on what? How big was each house supposed to be? I see you spent $750 on March 12th... what were you doing on the house build on that day? So, think that through from your business process.

Next. look for all the data available. Specifically, first, you need some kind of estimate of project scope. It doesn't matter whether the scope was correct, but it has to be the prior estimate, because you won't have a final estimate on anything you are predicting.

In the case of projects, maybe these will be categories as well. Work on SQL refactoring may be different from work on UI, or greenfield system design.

You also need to know how many resources were involved in producing that many SLOC. The SLOC didn't produce themselves.

Once you've gathered whatever data you have, figure out how to chunk it up.

If you are trying to predict completion, then think in terms of how you would relate the scope to the actual final SLOC invested. What may be the cool thing to apply ML to is identifying, given all this information, which projects are going sideways. Is there some common pattern to the projects that end up 200% above scope, as opposed to the normal ones that only were 130%?


By the way, PLEASE get in the habit of using ISO format for dates.

2018-01-03 is unambiguous. whereas 01.03.18 could be January 3rd or March 1st 2018...or even 1918, for that matter.

0 Karma
Get Updates on the Splunk Community!

Index This | I am a number, but when you add ‘G’ to me, I go away. What number am I?

March 2024 Edition Hayyy Splunk Education Enthusiasts and the Eternally Curious!  We’re back with another ...

What’s New in Splunk App for PCI Compliance 5.3.1?

The Splunk App for PCI Compliance allows customers to extend the power of their existing Splunk solution with ...

Extending Observability Content to Splunk Cloud

Register to join us !   In this Extending Observability Content to Splunk Cloud Tech Talk, you'll see how to ...