1

Suppose I have several points on a two dimensional plot. Is there an online program (or open source or free tool) that can get the most accurate formula for my plotted points? Or if I were to do this in Excel, what is the easiest way to figure out the graphing formula? I actually have plots related to record counts in a star schema fact table versus disk space growth in SQL Server, and I want to predict how much growth there will be in the next 12 months. Our clustered index is causing our growth to be exponential, so that is the reason I want to figure out the formula. The quick sorts for this clustered index also has to be factored in. Sometimes a clustered index can take up more space than the data because of the quick sorts, and needs to be allocated as temporary, but available, disk space as part of this calculation. If we weren't using a clustered index, I'd just use a simple ratio.

As a test, use these points for an example. I will get the disk space for each later.

134,000,000 (Jan 2013) > 301,000,000 (May 2013) > 1,696,000,000 (Aug 2013)

JustBeingHelpful
  • 442
  • 4
  • 16
  • There are infinitely many functions that could generate your points, so you need to constraint your model space, that is, you need to state what kind of functions you would like to approximate with, for example, step functions, picewise-linear, splines, some yet other approach? – dtldarek Jul 22 '13 at 08:52
  • I'm not a mathematician, so I don't know what function to start with. Given the 3 points I have so far, what is the simplest way to approach this and determine a graph function? I don't need a high level of accuracy. – JustBeingHelpful Jul 22 '13 at 17:52
  • If you suspect that your formula will be exponential, then first I would take a logarithm of all the data points and then I would approximate it with linear function or maybe some polynomial (if linear function fits very bad). – dtldarek Jul 22 '13 at 17:56
  • 2
  • Thanks so much. I recognize that site from a TED Talk I watched months ago. The last time I took a math course was 2004, so I'm a little rusty. – JustBeingHelpful Jul 22 '13 at 18:05

0 Answers0