Page 37 - Proceeding The 2nd International Seminar of Science and Technology : Accelerating Sustainable Innovation Towards Society 5.0
P. 37
nd
The 2 International Seminar of Science and Technology
“Accelerating Sustainable innovation towards Society 5.0”
ISST 2022 FST UT 2022
Universitas Terbuka
The analysis stages using the Random Forest method are as follows:
1. The first stage in the Random Forest Analysis is to input quarterly
data on gold jewellery demand from 2010 to 2021 into the R
Software.
2. Splitting data into training and testing data. Identification of the
Random Forest model with a predetermined ntree value (the
number of trees) using training data, data testing is used to see
the error rate of the model made.
3. Identify the model until you get the model with the smallest error.
4. Evaluation of the model and forecasting gold jewellery demand
for data in the sample and for the next year's data using the best
model with the smallest error.
2.2 Exponential Smoothing
Exponential Smoothing method is a moving average forecasting
technique with weights where the data is weighted by an exponential
function. Exponential smoothing is a moving average forecasting
method with advanced weighting, but still easy to use [12]. This
method has very little recording of past data or in other words this
method pays more attention to the value of the most recent
observations [7]. The analysis stages using the Single Exponential
Smoothing method are as follows:
1. The first stage in the Single Exponential Smoothing Analysis is to
input quarterly data on demand for gold jewellery from 2010 to
2021 into R Software.
2. The SES method uses data with a stationary pattern. A data can
be said to be stationary if there is no growth, decline, and the data
pattern is around a fixed average value ( ) = and the
2
variance around the average remains ( ) = ( − ) =
for a certain time [13].
2
3. It takes an alpha value, then it is tested for several alpha values,
namely (alpha = 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9) and the
smallest error value is obtained using an alpha value of 0.4.
4. Forecasting the amount of gold demand for data in the sample
and data for the next period using the best model (alpha = 0.4).
16 ISST 2022 – FST Universitas Terbuka, Indonesia
International Seminar of Science and Technology “Accelerating Sustainable
Towards Society 5.0