This is the time series analysis of Tajikistan local currency somoni (TJS). Here I used several famous predictions models such as Holt-Winters, ARIMA, SARIMAX and etc. Then I compare each model.
The dataset used is the National Bank of Tajikistan rate of USD/TJS dataset. This dataset contains rate of currencies data from 2001 to 2022 sampled everyday. I decided to resample the dataset with monthly frequency for both easier data handling and proximity to a real use case scenario (because NBT raises rate monthly, quaterly, semi-annually). In this case the series is not stationary with some small seasonalities(seasonal) which change every year(quaterly).
There are several popular models left, like LSTM, XGBoost, Prophet, Lasso and etc. that I should implement.
- Root Mean Squared Error (RMSE)
- Mean Absolute Percentage Error (MAPE)
Model | rmse | mape |
---|---|---|
SARIMA | 0.48 | 0.04 |
Holt-Winters | 0.41 | 0.03 |