ARIMA Equation Interpretation

Feb 19, 2008

Hello!

In AS 2008 it is possible to build time series model using ARIMA algorithm. But how exactly the ARIMA model eqation can be interpreted? For example, I have the following ARIMA equation in the mining legend of the (only) node:
ARIMA equation:
ARIMA ({1,1.06331498820371E-02,-9.1320555134096E-04,6.82290767622866E-03,5.16804012070969E-02,5.33387901298971E-02,7.27630334027759E-04,-4.08433722450516E-02},1,{1,0.637381531245194}) X ({1,8.85242031097984E-02},0,{1,3.46939494788227E-02})(26) X ({1},1,{1,-0.94291566648386})(13) Intercept:706.18948950439

I am used to the notations like ARIMA(Autoregressive order, Order of differencing, Moving average order). So, I can guess, that term like {1,1.06331498820371E-02,-9.1320555134096E-04,6.82290767622866E-03,5.16804012070969E-02,5.33387901298971E-02,7.27630334027759E-04,-4.08433722450516E-02} expresses coefficients in either autoregressive or moving average terms. But how can I interpret patterns followed by X like ARIMA(...)X(...)X(...)? Do they express periodical patterns? If so, how can I pick periods from this equation?

Thank you very much in advance.

View 3 Replies


ADVERTISEMENT

Data Mining :: ARIMA Equation Intercept Interpretation

Sep 6, 2015

I get an ARIMA equation in a form like below:

ARIMA ({1,-6.88479117532726E-02,-0.41430691637436},0,{1,4.42986549117398E-03}) X ({1,-0.473450289323228},0,{1,0.013772719862356,-2.14724317829266E-02,-4.80711350511344E-02,-1.15173259713497E-02,-9.86142925337024E-03,1.06803677786174})(6)
Intercept:18561.1721123981

I do understand the coefficients part and the seasonal part as well, however, I have trouble deciphering what exactly is the intercept here. Initially, I thought it would be equivalent to the value of mu in the original Box-Jenkins equation (so indicate a level), then I thought it's an intercept for the AR terms. It seems that none of these interpretations is good. The only official explanation I found at msdn was:

"Normally, the intercept (VALUETYPE = 11) or residual in a regression equation tells you the value of the predictable attribute, at the point where the input attribute, is 0. In many cases, this might not happen, and could lead to counterintuitive results.

For example, in a model that predicts income based on age, it is useless to learn the income at age 0. In real-life, it is typically more useful to know about the behavior of the line with respect to average values. Therefore, SQL Server Analysis Services modifies the intercept to express each regressor in a relationship with the mean.

This adjustment is difficult to see in the mining model content, but is apparent if you view the completed equation in the Mining Legend of the Microsoft Tree Viewer. The regression formula is shifted away from the 0 point to the point that represents the mean. This presents a view that is more intuitive given the current data.Therefore, assuming that the mean age is around 45, the intercept (VALUETYPE = 11) for the regression formula tells you the mean income.

View 2 Replies View Related

Rank Equation

Jun 7, 2006

Hi all,

until recently ive been using a rank equation to calculate rank,

essentially doing a select statement and selecting this as the rank field, where query 2 is the same as query 1:

((select count(*) from (query1) where (query1).value < (query2).value) +1)) as Rank

problem is that this is now running like a dog (takes 10 secs) and i'd like to try and do this another way- 2005 has a rank function, how can i do this in 2000?

here is the full statement :

SELECT StudentId, GCSE_Score, LTRIM(STR
((SELECT COUNT(*)
FROM dbo.[Score2004-05]
WHERE GCSE_score < s.GCSE_Score) + 1)) AS GCSE_Rank, SetId
FROM dbo.[Score2004-05] S
WHERE (SetId = '2004/2005')

greg

View 4 Replies View Related

Summing An Equation

Jun 6, 2008

This is a working 12 month intrest equation. I used this for the layout section but I am trying to take this and it gives me the correct values. But what I need to do next is have it sum those values.

I tried =SUM( whole expression but that didnt work) you can laugh at me I know but any help would be great!

=Switch(Fields!eqprecdt.Value< CDate("1 Jan 2007"),Fields!bookvalue.Value*datediff("d",Now(),#1/1/2007#)* .07/365,Fields!eqprecdt.Value> CDate("1 Jan 2007"), Fields!bookvalue.Value * datediff("d",Now(),Fields!eqprecdt.Value)* .07/365)*-1

View 5 Replies View Related

Does HTML Interpretation In SQL Server Reporting Services Exist?

Apr 30, 2007

Hi everybody



Is there a way to show HTML tags in a SQL Server Reporting services just like we are used to do in Crystal Reports?




I my case, I use the property Text Interpretation: HTML (right click in parameter field inside report and click edit parameter menu option).




I was wondering if we have something similiar in Reporting services?!

View 6 Replies View Related

Very Complex Equation Using TSQL Need Help

Sep 25, 2007

I am not sure if I am posting this is the right place, please let me know if a better place exists.I am trying to figure out the distance from a point on the earth to a route between two other points. All points that I have are in latitude and longitude secondsIE: 51-52-40.6700N => 186760.6700NI am able to find the distance between two points on the globe using a custom function FindDist(latitude1, longitude1, latitude2, longitude2)I did the legwork for the math and was able to come up with this psudo code: p1 = start of routep2 = end of routep3 = point of referencep4 = INTERSECT(LINE(p1,p2), PERPENDICULAR_LINE(p1,p2,p3)) The direct distance between p4 and p3 is somewhere in the range of 0 - 70 and encompasses all points in the US (We only deal in the US) which is incorrect. If I do the conversion of p4 to latitude and longitude I get numbers that start out at 1000 which is also incorrect.I am doing the conversion with the following: DECLARE @lat2 VARCHAR(25)SET @lat2 = (@p4x * 3600)IF @p4x < 0BEGINSET @lat2 = @lat2 + 'S'
ENDIF @p4x > 0BEGINSET @lat2 = @lat2 + 'N'
ENDDECLARE @lng2 VARCHAR(25)SET @lng2 = @p4y * 3600IF @p4y < 0BEGINSET @lng2 = @lng2 + 'W'
ENDIF @p4y > 0BEGINSET @lng2 = @lng2 + 'E'
END I am not sure, however, how to either:a) find the latitude / longitude of p4b) find the distance in miles between p3 and p4Either of these will solve my problemI know this is a rather odd question, but if anyone can help me that would be great. If you need more information, let me know and I will try and explain things further. 

View 1 Replies View Related

Time Series In SQL Server 2005 Desicion Tree Interpretation

Aug 12, 2007



Dear All
The problem is I am going to predict the production for different category of product.
attributes are
year - key
A production - predict only
B production -predict only
C production -predict only

And in the SQL it is impossible to to give input and predict (I am not sure whether that is a error or not).

And in the decision tree
for the
Product A - get as product A >=12324
Product B year > 2000
Product C product C >=35454


I want to know why the label is changing time to time.
Please help me on this. Thank you
Menik



View 3 Replies View Related

How To Pass Result From Case Statement To An Equation?

Dec 28, 2007

How to get the CASE results highlighted in BOLD into this equation; "(LogOut - LogIn) + (LunchBreak) -(AMBreak)  + (PMBreak)  AS TimeWorked" ?
Thank you.
CREATE VIEW dbo.vwu_ReportViewASSELECT   EmployeeID  , LastName  , FirstName  , LocationCode  , UserID  , Today  , Login  , AMBreakOut  , AMBreakIn            , CASE  WHEN ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) >= 0 AND ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) <= 19    THEN '0'  WHEN ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) >= 20 AND ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) <= 34    THEN '15'  WHEN ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) >= 35 AND ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) <= 49    THEN '30'  WHEN ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) > = 50 AND ISNULL(DATEDIFF(Minute, AMBreakOut, AMBreakIn),0) <= 64    THEN '45'  ELSE '60' END AS AMBreak , LunchOut             , LunchIn            , CASE  WHEN ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0)  >= 0 AND ISNULL(DATEDIFF(Minute,  LunchOut, LunchIn),0) <= 66    THEN '0'  WHEN ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0)  >= 67 AND ISNULL(DATEDIFF(Minute,  LunchOut, LunchIn),0) <= 81    THEN '15'  WHEN ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0)  >= 82 AND ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0) <= 96    THEN '30'  WHEN ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0)  >= 97  AND ISNULL(DATEDIFF(Minute,  LunchOut, LunchIn),0) <= 111    THEN '45'  WHEN ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0)  >= 112 AND ISNULL(DATEDIFF(Minute, LunchOut, LunchIn),0) <= 126    THEN '60'  ELSE '75' END AS LunchBreak, PMBreakOut , PMBreakIn            , CASE  WHEN ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) >= 0 AND ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) <= 19    THEN '0'  WHEN ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) >= 20 AND ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) <= 34    THEN '15'  WHEN ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) >= 35 AND ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) <= 49    THEN '30'  WHEN ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) >= 50 AND ISNULL(DATEDIFF(Minute, PMBreakOut, PMBreakIn),0) <= 64    THEN '45'  ELSE '60' END AS PMBreak , Logout , Comments , LoginLogon , AMBreakOutLogon , AMBreakInLogon , LunchOutLogon , LunchInLogon , PMBreakOutLogon , PMBreakInLogon , LogoutLogon             ,(LogOut - LogIn) + (LunchBreak) -(AMBreak)  + (PMBreak)  AS TimeWorked 

View 7 Replies View Related







Copyrights 2005-15 www.BigResource.com, All rights reserved