Putri, Talitha Adilla Fujisai Panglima (2026) Peramalan Inflasi Berbasis Deep Learning dengan Variabel Makroekonomi dan Efek Kalender di Indonesia Menggunakan N-BEATSx dan Bayesian Optimization. Undergraduate thesis, UPN Veteran Jawa Timur.
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Cover)
22083010079.-cover.pdf Download (1MB) |
|
|
Text (Bab 1)
22083010079.-bab1.pdf Download (359kB) | Preview |
|
|
Text (Bab 2)
22083010079.-bab2.pdf Restricted to Repository staff only until 9 July 2029. Download (680kB) |
|
|
Text (Bab 3)
22083010079.-bab3.pdf Restricted to Repository staff only until 9 July 2029. Download (941kB) |
|
|
Text (Bab 4)
22083010079.-bab4.pdf Restricted to Repository staff only until 9 July 2029. Download (2MB) |
|
|
Text (Bab 5)
22083010079.-bab5.pdf Download (229kB) | Preview |
|
|
Text (Daftar Pustaka)
22083010079.-daftarpustaka.pdf Download (203kB) | Preview |
|
|
Text (Lampiran)
22083010079.-lampiran.pdf Restricted to Repository staff only until 9 July 2029. Download (249kB) |
Abstract
Inflation is a key indicator of economic stability that plays a crucial role in formulating monetary and fiscal policies. In Indonesia, inflation patterns are influenced not only by macroeconomic factors but also by seasonal variations resulting from religious holidays, which cause periodic price fluctuations. Conventional forecasting methods generally have limitations in capturing nonlinear relationships and the simultaneous effects of exogenous variables and calendar effects. This study develops an inflation forecasting model using N-BEATSx, a decomposition-based deep learning model that captures trend and seasonal components and integrates macroeconomic exogenous variables and calendar effects into the forecasting process. The data used consists of monthly time series from January 2009 to September 2025, covering inflation, the BI Rate, the USD/IDR exchange rate, global oil prices, lagged inflation variables, and calendar dummies for Ramadan, Eid al-Fitr, Christmas, and the Lunar New Year. Hyperparameter optimization was performed via two-stage Bayesian Optimization using Optuna. The model was evaluated using MAE, RMSE, and SMAPE, and compared with the Prophet, SARIMAX, and LSTM models. Results show that N-BEATSx with Bayesian Optimization yields an MAE of 0.00601, RMSE of 0.00834, and SMAPE of 41.76%, outperforming Prophet (MAE=0.00487, RMSE=0.00592, SMAPE=43.96%), SARIMAX (MAE=0.00717, RMSE=0.00905, SMAPE=46.40%), and LSTM (MAE=0.00964, RMSE=0.01149, SMAPE=58.30%) in cross-period prediction consistency. Decomposition analysis indicates that exogenous variables contribute an average of 59.60% to the forecast, highlighting the importance of macroeconomic and calendar factors in shaping inflation. The six-month inflation forecast indicates a range of 2.41% - 2.78%, which can serve as a reference for policymakers in anticipating short-term inflation dynamics in Indonesia.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | H Social Sciences > HA Statistics | ||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | Miss Talitha Panglima | ||||||||||||
| Date Deposited: | 09 Jul 2026 04:38 | ||||||||||||
| Last Modified: | 09 Jul 2026 04:38 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/54785 |
Actions (login required)
 |
View Item |