1
$\begingroup$

I have sales data of my organization for last 3 years and based on that I have to forecast Budget amount for next Fiscal Year. Sales data is available across different Geographies for various Product subscription.

Now, I have gone through some blogs related to different Time series modeling on Analytics Vidhya and MachineLearningMastery and understand the basics pretty well.

Que:

  1. What is the ideal way to approach this problem?
  2. Should I use Time series analysis(ARIMA) or simple regression would be the best option.

Acc. to me since I have cross-sectional data(across geo and products) regression would best fit my problem. Though, if I divide my data into Distinct different combinations(Let's say 100) of Product/Geo/Year/Months, I might end up having 100 diff series which can be forecasted using ARIMA.

Any help/suggestion/reference are appreciated.

Regards, Jagbir

$\endgroup$
1
$\begingroup$

There is no way to know what will have the highest predictive accuracy, thus the typical solution is to try both on multiple test sets and see which performs best. Using regression only will leave open the possibility of unaccounted for, predictable error such as autocorrelations. Vector Autoregressions might also be a consideration. You can also try using exponential smoothing with regressors with some packages (such as 'smooth' in R).

$\endgroup$
0
$\begingroup$

To better understand the approach ....review Time Series Analysis (memory) vs Regression (causal) here http://www.autobox.com/pdfs/regvsbox-old.pdf and merge the two to develop a SARMAX model http://www.autobox.com/pdfs/SARMAX.pdf which might include both.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.