Compare between the different approaches of forecasting and advise by return the one you suggest

Jason000000

New member
Two automotive companies are trying to forecast the next year sales. They try to select the best approach and tool to make the forecast as accurate as possible. Compare between the different approaches of forecasting and advise by return the one you suggest, and mention why did not you use the other

Year 1st company sales ( In millions) Month 2nd company sales( In millions)
2010 22 October 2017 42
2011 11 November 2017 47
2012 35 December 2017 37
2013 49
2014 52
2015 46
2017 50

Jameson

Staff member
Two automotive companies are trying to forecast the next year sales. They try to select the best approach and tool to make the forecast as accurate as possible. Compare between the different approaches of forecasting and advise by return the one you suggest, and mention why did not you use the other

Year 1st company sales ( In millions) Month 2nd company sales( In millions)
2010 22 October 2017 42
2011 11 November 2017 47
2012 35 December 2017 37
2013 49
2014 52
2015 46
2017 50
Hi Jason.

So this is a time series problem most likely. It could be modeled in other ways but usually when we group by calendar month it involves time series. What topic(s) have you covered around this topic? What have you tried? Without some context it's hard to give guidance.

Jason000000

New member
Hi Jason.

So this is a time series problem most likely. It could be modeled in other ways but usually when we group by calendar month it involves time series. What topic(s) have you covered around this topic? What have you tried? Without some context it's hard to give guidance.
Hi Jameson,

yes you are right it's a time series method .. one of many forecast technique ... and my guess it's the seasonality pattern of time series.
The problem is how to apply this seasonal pattern on both companies!

Year................1st company sales ( In millions)
2010......................................22
2011......................................11
2012......................................35
2013......................................49
2014......................................52
2015......................................46
2016......................................48
2017......................................50

Month......................2nd company sales( In millions)
October 2017..........................42
November 2017.......................47
December 2017.......................37

Jason000000

New member
Hi Jameson,

yes you are right it's a time series method .. one of many forecast technique ... and my guess it's the seasonality pattern of time series.
The problem is how to apply this seasonal pattern on both companies!

Year................1st company sales ( In millions)
2010......................................22
2011......................................11
2012......................................35
2013......................................49
2014......................................52
2015......................................46
2016......................................48
2017......................................50

Month......................2nd company sales( In millions)
October 2017..........................42
November 2017.......................47
December 2017.......................37

my suggestion is to apply the "moving average" method of forecasting on company 1, and get the usual MAD, MSE & MAPE.
and as for the 2nd company.. due to very limited records I suggest applying the "naive" method. And also get the MAD, MSE & MAPE.

What do you think?

Jameson

Staff member
my suggestion is to apply the "moving average" method of forecasting on company 1, and get the usual MAD, MSE & MAPE.
and as for the 2nd company.. due to very limited records I suggest applying the "naive" method. And also get the MAD, MSE & MAPE.

What do you think?
Hi Jason,

This is data set is very, very small. Usually for time series we have a minimum of 30 points and ideally more like 100. So I think moving average is reasonable given this constraint. All of those metrics are fine to use. If this is for a course I'm really surprised by the lack of data. In my job we come across this issue sometimes but in a classroom they should try to construct usable data sets. Anyway, what do you get for the moving average? Over how many points do you propose to average?

Jason000000

New member
Hi Jason,

This is data set is very, very small. Usually for time series we have a minimum of 30 points and ideally more like 100. So I think moving average is reasonable given this constraint. All of those metrics are fine to use. If this is for a course I'm really surprised by the lack of data. In my job we come across this issue sometimes but in a classroom they should try to construct usable data sets. Anyway, what do you get for the moving average? Over how many points do you propose to average?
Hey Jameson .. thanx for the input .. i will pass you what I achieved to check it out .. thanx

Last edited: