| View previous topic :: View next topic |
| Author |
Message |
passmaster16
Newbie
Joined: Mar 13, 2011
Posts: 2
|
 Posted: Mon Mar 14, 2011 9:49 am Post subject: Trend Analysis Tools? |
 |
|
We're looking to implement problem management into our environment. We currently use Frontrange Heat for incident management. From our Heat system we have various reports such as top 10 call types in a period. Our call types and subtypes are fairly well defined. We also dump the data into excel and use pivot tables to give us a better representation of the breakdown of the incidents.
With that being said, we can use the current reporting tools to see changes from period to period, but I was wondering if there are any tools out there that could parse data perhaps based on keywords to find similarities between various incident calls. As it stands right now, we would have to manually go through the call descriptions to see if any correlation exists among the incidents. It would be nice to have a tool that could analyze the dataset and group any incidents that could be related to each other. Any thoughts on this?
Thanks! |
|
| Back to top |
|
 |
Bluesman
Itiler
Joined: Mar 19, 2008
Posts: 29
|
| Posted: Mon Mar 14, 2011 9:41 pm Post subject: |
|
|
Hmm. Frontrange runs a database (SQL server is one option, IIRC). That data can be queried by a few SQL Server tools, if the Heat tool doesn´t contain those possibilities. Check the help for Heat first ( on writing SQL queries directly from Heat). Then start looking at SQL server Analysis Services. From there, I´d say you have a good chance of making trends/prognosis work that will actually be good.
Database diagram is necessary, of course. And an intimate knowhow of Transact-SQL and Heat.
Good luck!
_________________
---------------- bragging line -----------------
V2 Combined Practitioner (service support)
Problem Mgr |
|
| Back to top |
|
|
BorisBear
Senior Itiler
Joined: Mar 10, 2008
Posts: 403
Location: Sunderland
|
| Posted: Tue Mar 15, 2011 1:33 am Post subject: |
|
|
Sounds a bit rubbish to me. Your database fields should be meaningful enough to allow effective trend analysis and their population should be rigidly adhered to by users of the system.
What you're proposing is performing analytics based on specific rules against a wide, inconsistent data set
Trend Analysis often suffers from the SISO issue - Sh!te In Sh!te Out |
|
| Back to top |
|
|
passmaster16
Newbie
Joined: Mar 13, 2011
Posts: 2
|
| Posted: Tue Mar 15, 2011 7:56 am Post subject: |
|
|
| BorisBear wrote: | Sounds a bit rubbish to me. Your database fields should be meaningful enough to allow effective trend analysis and their population should be rigidly adhered to by users of the system.
What you're proposing is performing analytics based on specific rules against a wide, inconsistent data set
Trend Analysis often suffers from the SISO issue - Sh!te In Sh!te Out |
Then we agree to disagree. We are breaking our call types down to be more granular, and that will certainly help with trend analysis. But at some point, somebody would have to analyze the data if there was a fluctuation from month to month to see if there was any trend that could be correlated.
I'm not asking to perform analysis based on specific rules. I'm asking if there is a tool that could examine a dataset and find similarities between the records. At that point I could examine the output to see if it's legitimate or not.
And Bluesman is right, our system runs on an Oracle db so we could run sql scripts against it. I just wanted to stop in here first to see if anybody knew of any tools before we do something custom.
Thanks for your input. |
|
| Back to top |
|
|
BorisBear
Senior Itiler
Joined: Mar 10, 2008
Posts: 403
Location: Sunderland
|
| Posted: Tue Mar 15, 2011 7:23 pm Post subject: |
|
|
| passmaster16 wrote: | | BorisBear wrote: | Sounds a bit rubbish to me. Your database fields should be meaningful enough to allow effective trend analysis and their population should be rigidly adhered to by users of the system.
What you're proposing is performing analytics based on specific rules against a wide, inconsistent data set
Trend Analysis often suffers from the SISO issue - Sh!te In Sh!te Out |
Then we agree to disagree. We are breaking our call types down to be more granular, and that will certainly help with trend analysis. But at some point, somebody would have to analyze the data if there was a fluctuation from month to month to see if there was any trend that could be correlated.
I'm not asking to perform analysis based on specific rules. I'm asking if there is a tool that could examine a dataset and find similarities between the records. At that point I could examine the output to see if it's legitimate or not.
And Bluesman is right, our system runs on an Oracle db so we could run sql scripts against it. I just wanted to stop in here first to see if anybody knew of any tools before we do something custom.
Thanks for your input. |
Whatever you pull from SQL you're going to have to analyse further........and how confident can you be that it's comprehensive? How many different ways do we have of describing the same thing? So long as you walk into it knowing that it's only a small part of the story that's ok.......further manual effort may be required. Maybe it's easier to perform analytics to establish where good information hasn't been input (the infamous 'Other' category) and tackle the users of your tool.
So have you already done all the Pareto analysis and garnered all of the stakeholder perception about where biggest the problems lie? |
|
| Back to top |
|
|
BorisBear
Senior Itiler
Joined: Mar 10, 2008
Posts: 403
Location: Sunderland
|
| Posted: Tue Mar 15, 2011 9:20 pm Post subject: |
|
|
Just a further thought......have you considered using collocation software?
There is a lot of freeware/optionware out there that may help you |
|
| Back to top |
|
|
Diarmid
Senior Itiler
Joined: Mar 04, 2008
Posts: 1884
Location: Newcastle-under-Lyme
|
| Posted: Wed Mar 16, 2011 11:25 pm Post subject: |
|
|
The real problem with attempting to automatically match free text is not the false links (which can be readily dismissed by further - manual - examination), but the difficulty in remembering that it might be missing important associations which fall outside the rules you set up, i.e. of assuming you need look no further.
I believe that some organizations take the step of opening a problem investigation for any incident that is not matched to an already identified problem. Some people believe that is the only "good" approach.
_________________
"Method goes far to prevent trouble in business: for it makes the task easy, hinders confusion, saves abundance of time, and instructs those that have business depending, both what to do and what to hope."
William Penn 1644-1718 |
|
| Back to top |
|
|
BorisBear
Senior Itiler
Joined: Mar 10, 2008
Posts: 403
Location: Sunderland
|
| Posted: Thu Mar 17, 2011 9:16 pm Post subject: |
|
|
| Diarmid wrote: | The real problem with attempting to automatically match free text is not the false links (which can be readily dismissed by further - manual - examination), but the difficulty in remembering that it might be missing important associations which fall outside the rules you set up, i.e. of assuming you need look no further.
I believe that some organizations take the step of opening a problem investigation for any incident that is not matched to an already identified problem. Some people believe that is the only "good" approach. |
Expressed far more articulately than I managed 
Treated as being a representative sample its fine so long as you understand the confidence rating. As I said previously I would maybe use it to reinforce correct usage of defined categories and correct any misuse of the ITSM tool. |
|
| Back to top |
|
|
|
Forum FAQ
Search
Usergroups
Profile
Log in to check your private messages
Log in
