DYNAMIC FACTOR MODEL ANALYSIS BASED ON PRINCIPAL COMPONENT EXTRACTED FACTOR FOR GRDP PREDICTION

AGUSTINO, SIMATUPANG (2026) DYNAMIC FACTOR MODEL ANALYSIS BASED ON PRINCIPAL COMPONENT EXTRACTED FACTOR FOR GRDP PREDICTION. FAKULTAS MATEMATIKA DAN ILMU PENGETAHUAN ALAM, UNIVERSITAS LAMPUNG.

[img]
Preview
 Text
ABSTRAK.pdf
Download (253kB) | Preview
[img] Text
SKRIPSI FULL.pdf
Restricted to Repository staff only
Download (3MB) | Request a copy
[img]
Preview
 Text
SKRIPSI TANPA BAB PEMBAHASAN.pdf
Download (2MB) | Preview

Abstract

Dynamic Factor Model (DFM) merupakan model peramalan multivariat yang merepresentasikan pergerakan bersama antarvariabel melalui sejumlah kecil faktor laten. DFM unggul dalam mengekstraksi informasi bersama dari variabel yang memiliki korelasi tinggi serta mengurangi pengaruh idiosyncratic noise, sehingga meningkatkan akurasi peramalan. Penelitian ini bertujuan menerapkan DFM untuk memodelkan dan meramalkan Produk Domestik Regional Bruto (PDRB) riil dan nominal Provinsi Lampung. Keterbatasan data PDRB triwulanan diatasi melalui disagregasi temporal menggunakan metode Denton untuk memperoleh data bulanan. Selanjutnya, faktor laten diekstraksi menggunakan Principal Component Analysis (PCA) dan diestimasi dalam kerangka state-space menggunakan Kalman Filter dengan Maximum Likelihood Estimation (MLE). Spesifikasi AR(1), AR(2), dan AR(3) dibandingkan menggunakan Log-Likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), uji Ljung–Box, dan Mean Absolute Percentage Error (MAPE). Hasil menunjukkan bahwa komponen utama pertama menjelaskan sekitar 99% variasi data. Model DFM-AR(3) memberikan kinerja terbaik serta menghasilkan akurasi tertinggi dalam merekonstruksi data PDRB triwulanan, dengan nilai MAPE sebesar 1,43% untuk PDRB riil dan 1,52% untuk PDRB nominal. Temuan ini menunjukkan bahwa DFM merupakan pendekatan yang efektif untuk memodelkan dan meramalkan dinamika PDRB menggunakan variabel yang berkorelasi tinggi. Keywords: Dynamic Factor Model, Principal Component Analysis, Metode Denton, Peramalan, PDRB Dynamic Factor Model (DFM) are multivariate forecasting models that represent the co-movements of observed variables through a small number of latent factors. DFMs effectively extract common information from highly correlated variables while reducing the influence of idiosyncratic noise, thereby improving forecasting accuracy. This study applies DFM to model and forecast the real and nominal Gross Regional Domestic Product (GRDP) of Lampung Province. The limitation of quarterly GRDP data was addressed through Denton temporal disaggregation to obtain monthly observations. Latent factors were extracted using Principal Component Analysis (PCA) and estimated within a state-space framework using the Kalman Filter under the Maximum Likelihood Estimation (MLE) framework. The AR(1), AR(2), and AR(3) specifications were evaluated using the LogLikelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), the Ljung–Box test, and the Mean Absolute Percentage Error (MAPE). The first principal component explained approximately 99% of the total data variation. The DFM-AR(3) model achieved the best performance, reconstructing the original quarterly GRDP series with MAPE values of 1.43% for real GRDP and 1.52% for nominal GRDP. These results demonstrate that DFM is an effective approach for modeling and forecasting GRDP using a limited number of highly correlated variables. Keywords: Dynamic Factor Model, Principal Component Analysis, Denton Method, Forecasting, GRDP

Item Type: Other
Subjects: ?? 510 ??
Divisions: Fakultas MIPA > Prodi Matematika
Depositing User: 2606826815 Digilib
Date Deposited: 06 Jul 2026 04:18
Last Modified: 06 Jul 2026 04:18
URI: http://digilib.unila.ac.id/id/eprint/102070

Actions (login required)

View Item View Item