topic Re: point prediction in Python Questions

point prediction

MOHAMMEDHABBOUB — Wed, 27 Feb 2013 15:15:00 GMT

Hello everybody,

I have 7 points (x,y coordinate is known) each point represent a position of the same object in a specific time of the year (say January 1984, January 1986 ,January 1999.....) [See attached file]

How could I know the expected position in (say January 2014?) using python or Arcgis?

Yours
Habboub

Prediction a point position

MOHAMMEDHABBOUB — Wed, 27 Feb 2013 15:17:58 GMT

Re: point prediction

JimCousins — Wed, 27 Feb 2013 17:37:07 GMT

Habboub,
There must be some variables that impact the movement over time, and those need to be assigned weighting factors for use in the prediction of future locations. You need to understand why 1984 to 1986 2 year period is comparatively large vs 1992 to 1999 7 year period. Any prediction based on the xy locations versus time without consideration of other variables may as well be random.
Try adding some more information for better responses.
Regards,
Jim

Re: point prediction

MOHAMMEDHABBOUB — Wed, 27 Feb 2013 17:43:02 GMT

Habboub,
There must be some variables that impact the movement over time, and those need to be assigned weighting factors for use in the prediction of future locations. You need to understand why 1984 to 1986 2 year period is comparatively large vs 1992 to 1999 7 year period. Any prediction based on the xy locations versus time without consideration of other variables may as well be random.
Try adding some more information for better responses.
Regards,
Jim

Well, If I add a rainfall as the main parameter>>> could this be done?

Re: point prediction

DarrenWiens2 — Wed, 27 Feb 2013 17:58:04 GMT

I realize the original post wants to do this in GIS, so if you already know how to do this outside of GIS, skip this comment. This is a regression problem, where your x variable varies over time (x in the graph). You can simply plot the points (in Excel, R, or whatever) and calculate the regression line to extrapolate outside the measured range. Find the equation for the regression line, and solve for your x. For example, in the attached graph, I made up y values: 1984 = 0, 1986 = 1, 1992 = 4.5, 1999 = 4.6, 2003 = 6, 2004 = 6.1, 2010 = 6.8 (obviously you have the correct x values). 2014 would have a value of 8.521.

Re: point prediction

MOHAMMEDHABBOUB — Thu, 28 Feb 2013 01:56:43 GMT

I realize the original post wants to do this in GIS, so if you already know how to do this outside of GIS, skip this comment. This is a regression problem, where your x variable varies over time (x in the graph). You can simply plot the points (in Excel, R, or whatever) and calculate the regression line to extrapolate outside the measured range. Find the equation for the regression line, and solve for your x. For example, in the attached graph, I made up y values: 1984 = 0, 1986 = 1, 1992 = 4.5, 1999 = 4.6, 2003 = 6, 2004 = 6.1, 2010 = 6.8 (obviously you have the correct x values). 2014 would have a value of 8.521.

Thank you, but as you noticed I was looking for a way to do so in arcgis or any other GIS software

Re: point prediction

ChrisSnyder — Thu, 28 Feb 2013 16:54:19 GMT

As far as I know, you can't directly perform linear regression (or other types of regression) in ArcGIS. You would have to:

1. Code the regression it yourself in Python or some other language that ArcGIS will interface directly with.
2. Use R or Excel, and then apply the derived equation using a Python cursor or some other data automation technique to put the new points in teh proper location.

What is the phenomenon you are trying to predict? If you gave more background on the topic, you might get more/better feedback. For example, is it expected that the the future point locations will always be to the east of the previous point? Can you expect the rate of change to increase (get faster) or decrease (get slower) over time?