Rahayu, Restu Sri (2024) Aplikasi Dynamic Factor Model dan Long Short Term Memory Berdasarkan Data Google Trends dan Makroekonomi untuk Nowcasting Pertumbuhan Produk Domestik Bruto Indonesia. Masters thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 6003211009-Master_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
6003211009-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (5MB) | Request a copy |
Abstract
Dynamic Factor Model (DFM) merupakan perkembangan dari analisis deret waktu yang digunakan untuk melakukan peramalan jarak dekat atau nowcasting. Model ini dapat menghubungkan variabel prediktor yang berfrekuensi bulanan dengan variabel respon yang berfrekuensi kuartalan. Selain DFM, salah satu algoritma yang sering digunakan untuk nowcasting adalah Long Short Term Memory (LSTM) karena kinerja baiknya dalam memprediksi nowcasting. DFM telah banyak digunakan di bidang ekonomi karena dapat menghubungkan variabel ekonomi yang diamati dalam kurun waktu yang berbeda. Ketersediaan data ekonomi yang cenderung mengalami keterlambatan merupakan salah satu masalah yang umumnya dijumpai. Secara umum, penyusunan kebijakan ekonomi memerlukan informasi kondisi ekonomi yang tersedia secara tepat waktu. Akan tetapi, indikator ekonomi makro cenderung dirilis dengan penundaan. Kondisi ekonomi suatu negara dapat tercemin dari Produk Domestik Bruto (PDB) negara tersebut. Laju Produk Domestik Bruto (PDB) memiliki peran yang sangat penting dalam membantu para pembuat kebijakan maupun pebisnis untuk mengerti kondisi perekonomian negara. Data PDB mengalami delay dalam perilisannya selama lima pekan sejak triwulan berakhir. Kondisi ini terjadi pada tingkat nasional dan regional. Oleh sebab itu, penelitian ini bertujuan melakukan nowcasting pertumbuhan atau laju PDB Indonesia dalam periode bulanan menggunakan data official statistics dan data google trends menggunakan Dynamic Factor Model (DFM) dan Long Short Term Memory (LSTM). Hasil pemodelan menunjukkan bahwa model DFM (r=6, p=3, q=5) memberikan hasil nowcasting pada data out-sample terbaik untuk studi kasus nowcasting laju pertumbuhan PDB Indonesia dibandingkan dengan metode LSTM. Model DFM ini memberikan hasil nowcasting dengan RMSE sebesar 1,873 dan sMAPE 0,932. Kata Kunci: Dynamic Factor Model (DFM), Long Short Term Memory (LSTM),nowcasting, PDB, Google Trends
========================================================================================================================
Dynamic Factor Model (DFM) is a development of time series analysis that is used to perform near-range forecasting or nowcasting. This model can relate the predictor variable with monthly frequency with the response variable with quarterly frequency. Besides DFM, one of the algorithms that is often used for nowcasting is Long Short Term Memory (LSTM) because of its good performance in predicting nowcasting. DFM has been widely used in economic fields because it can relate economic variables observed in different time periods. The availability of economic data, which tends to experience delays, is one of the problems that are commonly encountered. In general, formulating economic policies requires information on economic conditions that is available promptly. Policymakers need to know the current economic conditions as a foundation or a basis of projecting future economic conditions. However, macroeconomic indicators and the information needed tend to be released and available with a long delay. Furthermore, the economic condition of a country can be reflected in the country's Gross Domestic Product (GDP). The GDP’s growth is very important role in helping policy makers and business society understand the country's economic conditions. GDP data has been delayed in its release for five weeks since the quarter ended. This condition occurs at the national and regional levels. Therefore, this study aims to nowcast Indonesia's GDP growth or rate monthly using official statistics and google trends data using the Dynamic Factor Model (DFM) and Long Short Term Memory (LSTM). The modeling results show that the DFM model (r=6, p=3, q=5) provides the best nowcasting results on out-sample data for the nowcasting case study of Indonesia's GDP growth rate compared to the LSTM method. This DFM model provides nowcasting results with RMSE of 1,873 and sMAPE of 0,932. Key Words: Dynamic Factor Model (DFM), Long Short Term Memory (LSTM), nowcasting, GDP, Google Trends
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Dynamic Factor Model (DFM), Long Short Term Memory (LSTM),nowcasting, PDB, Google Trends,Dynamic Factor Model (DFM), Long Short Term Memory (LSTM), nowcasting, GDP, Google Trends. |
| Subjects: | H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HA Statistics > HA31.3 Regression. Correlation Q Science > QA Mathematics > QA278.5 Principal components analysis. Factor analysis. Correspondence analysis (Statistics) Q Science > QA Mathematics > QA280 Box-Jenkins forecasting |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis |
| Depositing User: | Restu Sri Rahayu |
| Date Deposited: | 05 Aug 2024 03:52 |
| Last Modified: | 05 Aug 2024 03:52 |
| URI: | http://repository.its.ac.id/id/eprint/113455 |
Actions (login required)
 |
View Item |