3

I have data in the interval of 5 years from 1975 to 2018 e.g. (1975-1979). (1980-1984) so on. I need to forecast the numbers in the future from 2018 to 2030 and 2040 in 5 years intervals also. what is the best method?

enter image description here

Tim
  • 108,699
  • 20
  • 212
  • 390
kirellos said
  • 31
  • 2

1 Answers1

2

You don't really any some special method. Time-series data always has a frequency, for example, it could be milliseconds, minutes, hours, weeks, or years. In your case, the frequency is 5-year periods. If you have 5-year data and you need to predict next 5 years, just train a time-series model on it and it will out-of-the-box make a prediction for the 5-year periods.

The only limitation is that your data comes in predefined boxes. So if you are at time $t$ equal to the 1975-79 box, the $t+1$ would be the 1980-84 box. If for example, you would need to make a prediction for the years 1977-81, you can use linear interpolation and take the appropriate fractions from both boxes $3/5 \times x_t + 2/5 \times x_{t+1}$.

Since your data is relatively small, you should look into simple forecasting methods that work well for short time-series.

Tim
  • 108,699
  • 20
  • 212
  • 390
  • can you send me a method for ordinary time series analysis? – kirellos said Sep 15 '21 at 08:52
  • @kirellossaid see edit. – Tim Sep 15 '21 at 08:55
  • 1
    For a great introduction to forecasting, I recommend [*Forecasting: Principles and Practice* (2nd ed.) by Athanasopoulos & Hyndman](https://otexts.org/fpp2/) and [*Forecasting: Principles and Practice* (3rd ed.) by Athanasopoulos & Hyndman](https://otexts.org/fpp3/) – Stephan Kolassa Sep 15 '21 at 10:00