A true I.E. Engineering/ Math Problem

[LBheat02]

I am faced with a project that I am truly stumped on. My objective is to take past historical parts data and use it to to forecast the demand monthly and yearly. This is just a sample of the data for parts sold per month with yearly total:

Year jan feb march april may june july....Total
2005 11 12 10 12 13 16 15.... 200
2006 10 12 9 12 12 15 10.... 210
2007 11 13 14 18 16 14 11.... 180

But the twist is that if we change the amount of parts we produce for a company per year. How do I make a program that will calculate forecast of demand if we change the total parts to produce next year to 220, based on the historical data i have stated. So inputs would be quantity of parts and another input would be how many months to calculate.

If you can help me on this i would truly appreciate it. I have to have this project done in 3 days and its a project for work lol. If you have any questions feel free to task at anytime.

[NotLKH]

Certainly,

This relates to some method that you have been exposed to within the previous few weeks/months at ?work?

Else

I would postulate that there isn't enough data to perform a monthly forcast.
And apparently the yearly forecast is irrelevent, since you are using a projection of 220 as an input for next year, ie.. a known.


BTW,

What did you graduate in?

-Lou

[jemidiah]

I'd figure you'd use some polynomial curve fit to do the forecasting--fit twelve curves, one for each month, to that month's data as a function of years. For various reasons this could be wildly inaccurate, but it's most likely fine. You'd want to graph the functions yourself to see how well they fit, and it would only work well if you have quite a few years' worth of data to go off of. Even a linear fit would probably be decent.

As for changing the number of parts to produce, that's simple enough. Forecast for the range and renormalize. That is, do your regular forecast for however many months, total how many parts you predict you should produce, and for each month multiply that month's prediction by (total number needed) / (total number forecasted). This will yield a total of (total number needed) parts.


Like Lou said, what did you get a degree in if you're having to do this kind of work seemingly without training?