AN ANALYSIS OF USING SVR AS AN ALTERNATIVE OF GARCH IN MODELING THE VOLATILITY OF JAKARTA COMPOSITE IDX INDEX
Volatility can be an impending risk while investing. The most common way to visualize the inherent risk of a financial instrument is using the ARCH-GARCH method. This method is not without flaws. This method is very susceptible to overfitting and cannot predict volatility in the future. To alleviate...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/60049 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Volatility can be an impending risk while investing. The most common way to visualize the inherent risk of a financial instrument is using the ARCH-GARCH method. This method is not without flaws. This method is very susceptible to overfitting and cannot predict volatility in the future. To alleviate the problem, GARCH can be replaced with a method that have more forecasting capabilities such as SVR. This study will use SVR to model the volatility of Jakarta IDX Composite Index from 27th of November 2020 until 4th February of 2021. This is achieved by modelling the return of the index with ARIMA (1,0) then using the GARCH and SVR method to model the volatility of said model. The SVR kernels used in this study is linear, polynomial, and RBF with a data partition of 1, 5, 10, 100, and 290. The best GARCH model that was used as a benchmark is GARCH (0,1) with a MAPE of 65,76%. After being examined qualitatively and quantitively examined, it is found that the linear and polynomial SVR models with 100 partitions can adequately model the volatility of said index. The models have a MAPE of 7.21% and 6.46% respectively. Lastly, both models did so without overfitting and still retains the predictive nature of SVR. |
|---|