I have the following dataset (.csv format) which contains:

  • 100 columns: $\textbf{Period}$ (in years, e.g. 2017, 2018, ..., 2028), $\textbf{Gross Revenue}$, $\textbf{Region}$ (e.g. APAC, NEMEA, etc), 97 other columns as $\textbf{factors}$ (e.g. # of customers, # of goods produced, etc).

  • 1,000,000 rows

where for $\textbf{Period}$, only 2017 is actual data, and 2018, ..., 2028 are $\textbf{planned}$, in the sense that data from these years are what the company expects to commit to and receive in these years in the future.

Now, I want to predict the gross sales for each $\textbf{Region}$ in 2018.

I tried using regression with stepwise, but the linear model that resulted from the stepwise had NA values for a majority of the predictors, and also the probability arising from the t-test.

I've consulted a few of my pals whom are working in the field of data science & analytics, and they all told me that it is impossible to predict for 2018..

Is it possible to do so given the state of the data? Some insight on this will be deeply appreciated!


1 Answer 1


Predicting next year's revenue from previous years' revenues is Time Series Forecasting. You will need data from many previous years to do this. One year is clearly not enough. Just think about it this way: Can you predict tomorrow stock's price given only today price?

  • $\begingroup$ Yes I absolutely agree on that. I'm currently working for a company, and am tasked on the above question, and got stuck at the model formulation. At this point, do you think I should advise my supervisor that it is impossible to perform such a prediction? Or if not, is it actually possible to predict next year's revenue based on planned revenues for the coming years? $\endgroup$
    – Stoner
    May 28, 2018 at 12:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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