SQL Clone
SQLServerCentral is supported by Redgate
 
Log in  ::  Register  ::  Not logged in
 
 
 


Forecasting with SQL


Forecasting with SQL

Author
Message
mark.wojciechowicz@gmail.com
mark.wojciechowicz@gmail.com
Valued Member
Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)

Group: General Forum Members
Points: 52 Visits: 588
Thanks for the clarifications, Al. I appreciate the finer points and I am sure this will be helpful to any readers as well.
Mark
piet_dj
piet_dj
Grasshopper
Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)Grasshopper (15 reputation)

Group: General Forum Members
Points: 15 Visits: 191
Do you have any useful info on calculating multiple linear regression with T-SQL?
SwePeso
SwePeso
SSCarpal Tunnel
SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)

Group: General Forum Members
Points: 4169 Visits: 3433
piet_dj (2/10/2010)
Do you have any useful info on calculating multiple linear regression with T-SQL?

Here http://developerworkshop.net/software.html
It also handles logarithmic, exponential and power regression.


N 56°04'39.16"
E 12°55'05.25"
ltaylor 73774
ltaylor 73774
Forum Newbie
Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)Forum Newbie (1 reputation)

Group: General Forum Members
Points: 1 Visits: 13
Very good article, thanks. When I picked up the code snippet and tried to use it, I saw one problem. In the last section of code, labeled "Create Forecast", the term that's multiplied by the seasonality factor is missing a parenthesis. That is,
MAX(A) + (MAX(B) * MAX(Forecastkey) + 1) should be
(MAX(A) + (MAX(B) * MAX(Forecastkey) + 1))

Correct me if I'm wrong, but this is the change that it took to make my forecast look correct.
mark.wojciechowicz@gmail.com
mark.wojciechowicz@gmail.com
Valued Member
Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)

Group: General Forum Members
Points: 52 Visits: 588
ltaylor 73774 (2/11/2010)
Very good article, thanks. When I picked up the code snippet and tried to use it, I saw one problem. In the last section of code, labeled "Create Forecast", the term that's multiplied by the seasonality factor is missing a parenthesis. That is,
MAX(A) + (MAX(B) * MAX(Forecastkey) + 1) should be
(MAX(A) + (MAX(B) * MAX(Forecastkey) + 1))

Correct me if I'm wrong, but this is the change that it took to make my forecast look correct.


Yes, you are correct. When the trend is multiplied against seasonality, the parenthese are needed.

Thank you for finding that error! I will get the download updated.

Mark
Tom Thomson
Tom Thomson
SSChampion
SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)SSChampion (14K reputation)

Group: General Forum Members
Points: 14550 Visits: 12238
Allen Nugent (2/9/2010)
Nice work, Mark.
I'll second that - a very nice article indeed.
Of course, there are weird and wonderful techniques for fitting data to nonlinear functions, but it might not be practical to do it with SQL.
Well, if you can do it say in Fortran+Dystal+LAPACK+FSPAK it's probably not too hard in SQL. What sort of package do people generally use for this sort of thing today? I'm way out of date (forgotten nearly all of my stats and most of my curve fitting) so I wouldn't try to do it myself. But if I asked an expert for predictions I would definitely want to know what methods he used, what assumptions he made, and maybe look at the code to check it didn't include the most common idiocies (and SQL is a language I can read, because it was designed for readability rather than to enable maximum obscurity like some pouplar languages).

Tom

Edward Boyle-478467
Edward Boyle-478467
SSC-Enthusiastic
SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)SSC-Enthusiastic (120 reputation)

Group: General Forum Members
Points: 120 Visits: 1152
Excellent article. I have a lot of forecast and actual time-series data and will check out the methods this weekend.
jolifox
jolifox
Grasshopper
Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)Grasshopper (16 reputation)

Group: General Forum Members
Points: 16 Visits: 5
Mark,

Very informative and useful article,

Do you have any suggestion on how to incorporate a weighting element into the SQL. i.e. if i wanted to give more weight to last years data over the year before etc.



mark.wojciechowicz@gmail.com
mark.wojciechowicz@gmail.com
Valued Member
Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)Valued Member (52 reputation)

Group: General Forum Members
Points: 52 Visits: 588
Hi Grasshopper,
I do not have any examples of weighted averages, but you may considering looking at exponential smoothing which weights, more heavily, recent periods.
I do suggest making sure that you prepare the data before forecasting on it. By using a moving median, you can eliminate a lot of the outliers.
Also, it's worth looking at the forecasting datamining algorithm in SSAS, which is a great tool.

Good luck,
Mark
SwePeso
SwePeso
SSCarpal Tunnel
SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)SSCarpal Tunnel (4.2K reputation)

Group: General Forum Members
Points: 4169 Visits: 3433
Here http://www.sqlteam.com/forums/topic.asp?TOPIC_ID=93911 is examples of both moving average, weighted average and weighted moving average.


N 56°04'39.16"
E 12°55'05.25"
Go


Permissions

You can't post new topics.
You can't post topic replies.
You can't post new polls.
You can't post replies to polls.
You can't edit your own topics.
You can't delete your own topics.
You can't edit other topics.
You can't delete other topics.
You can't edit your own posts.
You can't edit other posts.
You can't delete your own posts.
You can't delete other posts.
You can't post events.
You can't edit your own events.
You can't edit other events.
You can't delete your own events.
You can't delete other events.
You can't send private messages.
You can't send emails.
You can read topics.
You can't vote in polls.
You can't upload attachments.
You can download attachments.
You can't post HTML code.
You can't edit HTML code.
You can't post IFCode.
You can't post JavaScript.
You can post emoticons.
You can't post or upload images.

Select a forum

































































































































































SQLServerCentral


Search