Time Series Forecasting – Exponential Triple Smoothing

In an article dated 4 August 2017, a time series forecasting example based on ARIMA model was illustrated (Read more here).  Time series forecasting is very useful for business planning, especially on risk management in a systematic way.   However, not many companies have the resources to run ARIMA model using a statistical software.

Exponential Triple Smoothing (ETS), is another time series forecasting model that comes together with Microsoft Excel 2016 (Read more here).  It can be viewed as a subset of ARIMA model.  The ETS function in the Excel 2016 is very user friendly.  The forecast model can be generated within a few clicks.  The following chart shows the Malaysia Rubber Production forecast using the forecast function in Excel 2016.

The blue curve is the actual rubber production data taken from Bank Negara Malaysia (Read more here).  The orange curve is the forecasted data, while green and red curves are upper and lower bound of the forecasted data respectively.

The older version of Excel could only forecast using linear regression method.  The ETS algorithm in Excel 2016 is capable of studying the seasonal behaviour of the historical data and reflects them in the forecasted output.  The upper and lower confidence bound are basically two standard deviation from the mean.  In other words, we have 95% confidence level that the forecast prices for rubber production will lie 2.5% either way of the forecast.

From the above, it may be observed that for the years 2018 – 2020 (forecasted period), rubber production in Malaysia will continue to trend lower from decade ago, at a slower pace.  In addition, the seasonal lowest production months which usually happen during March-May is also “captured” by the ETS algorithm.