FORECASTING INFLATION IN INDONESIA USING ARIMAX-LIGHTGBM MODEL WITH OPTUNA HYPERPARAMETER OPTIMIZATION BASED ON MACROECONOMIC INDICATORS

Maharani, Dian (2026) FORECASTING INFLATION IN INDONESIA USING ARIMAX-LIGHTGBM MODEL WITH OPTUNA HYPERPARAMETER OPTIMIZATION BASED ON MACROECONOMIC INDICATORS. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010184_cover.pdf Download (1MB)
	Text (Bab 1) 22081010184_bab1.pdf Download (231kB)
	Text (Bab 2) 22081010184_bab2.pdf Restricted to Repository staff only until 12 July 2028. Download (635kB)
	Text (Bab 3) 22081010184_bab3.pdf Restricted to Repository staff only until 12 July 2028. Download (634kB)
	Text (Bab 4) 22081010184_bab4.pdf Restricted to Repository staff only until 12 July 2028. Download (2MB)
	Text (Bab 5) 22081010184_bab5.pdf Download (223kB)
	Text (Daftar Pustaka) 22081010184_daftarpustaka.pdf Download (193kB)

Abstract

Inflation is one of the key indicators reflecting a country’s economic stability; therefore, inflation forecasting is crucial as a basis for monetary policy-making and economic planning. However, the characteristics of inflation data, which contain linear and nonlinear patterns, mean that a single method is often unable to produce optimal forecasts. This study aims to develop an inflation forecasting model for Indonesia using a hybrid ARIMAX-LightGBM approach with Optuna hyperparameter optimization based on macroeconomic indicators. The exogenous variables used include money supply (M1 and M2), exchange rate, interest rates, inflation expectations, administered prices, and volatile food prices. The research stages include data cleaning, exploratory data analysis, data transformation, stationarity tests, ARIMAX model development, residual modeling using LightGBM based on lag features and statistics, hyperparameter optimization using Optuna, and evaluation using the Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Bias Error (MBE), and coefficient of determination (R²). The results show that the hybrid ARIMAX-LightGBM model with Optuna optimization delivers the best performance compared to the ARIMAX model, the LightGBM model, and the hybrid model without optimization. On the test data, the model produced an RMSE of 0.055973, an MAE of 0.042383, an MBE of 0.008892, and an R² of 0.989878. Furthermore, optimization using Optuna successfully improved the hybrid model’s performance by reducing the RMSE and MAE values compared to the model without optimization. Model analysis shows that information derived from the Administered Price and Volatile Food components contributed more significantly to improving inflation prediction capabilities than other macroeconomic variables during the study period.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Sari, Anggraini Puspita	NIDN0716088605	anggraini.puspita.if@upnjatim.ac.id
UNSPECIFIED	Al Haromainy, Muhammad Muharrom	NIDN0701069503	muhammad.muharrom.if@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Dian Maharani

Date Deposited:

13 Jul 2026 04:18

Last Modified:

13 Jul 2026 04:18

URI:

https://repository.upnjatim.ac.id/id/eprint/55183