Penerapan Model Hybrid Singular Spectrum Analysis – Time Series Regression – Neural Network Pada Peramalan Inflow Dan Outflow Uang Kartal Bank Indonesia

Balqis, Anisah (2024) Penerapan Model Hybrid Singular Spectrum Analysis – Time Series Regression – Neural Network Pada Peramalan Inflow Dan Outflow Uang Kartal Bank Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006201083-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (5MB) | Request a copy

Abstract

Sejarah perjalanan peradaban manusia menunjukkan bahwa uang memainkan peran penting dalam perekonomian, terutama sebagai alat pembayaran. Distribusi uang harus dikelola dengan bijak agar jumlah uang yang beredar sesuai dengan kebutuhan masyarakat. Bank Indonesia, sebagai lembaga yang memiliki wewenang untuk menjamin tersedianya uang rupiah yang layak edar perlu memiliki kemampuan untuk mengelola uang kartal dengan baik melalui perencanaan uang masuk dan uang keluar (inflow-outflow) dengan mempertimbangkan kebutuhan uang di masa depan dengan melakukan prediksi (forecasting). Penelitian ini bertujuan untuk memperoleh model peramalan inflow dan outflow di Indonesia, yaitu Nasional dan wilayah DKI Jakarta, yang akurat dan dapat menangkap pola variasi kalender. Untuk meningkatkan akurasi peramalan, digunakan metode neural network untuk menangkap pola nonlinier yang digabungkan dengan metode Singular Spectrum Analysis (SSA) dan Time Series Regression (TSR) menjadi hybrid SSA-TSR-NN. Metode hybrid SSA-TSR-NN akan dibandingkan dengan hybrid SSA-TSR-ARIMA berdasarkan RMSE, MAPE, dan MAE. SSA digunakan untuk memisahkan data deret waktu asli menjadi sejumlah komponen independen, termasuk komponen tren, musiman, dan noise. Metode TSR untuk memodelkan komponen tren dan komponen noise, sementara metode NN digunakan untuk memodelkan komponen musiman dan residual TSR komponen noise. Penelitian ini diaplikasikan pada dua wilayah data inflow dan outflow Bank Indonesia dari tahun Juli 2011 sampai Juni 2023. Kesimpulan yang diperoleh dari penelitian adalah secara umum, SSA-TSR-NN mampu mereduksi kesalahan peramalan SSA-TSR-ARIMA sebesar 27%. Neuron dan input yang digunakan pada NN bervariasi pada setiap data inflow dan outflow, jumlah neuron yang lebih banyak tidak selalu memengaruhi performa model menjadi lebih baik. Untuk peramalan satu tahun kedepan, inflow tertinggi jatuh pada bulan Januari dan April, sedangkan outflow tertinggi jatuh pada bulan Desember dan April.
==================================================================================================================================
The history of human civilization indicates that money plays a crucial role in the economy, particularly as a medium of exchange. The distribution of money needs to be managed wisely to ensure that the circulating amount aligns with societal needs. Bank Indonesia, as the institution authorized to guarantee the availability of viable Indonesian currency, needs the capability to effectively manage banknotes through the planning of inflow and outflow (inflow-outflow) of money. This involves considering future monetary needs through forecasting. This research aims to obtain accurate forecasting models for inflow and outflow in Indonesia, both at the national and DKI Jakarta regional levels, capturing calendar variation patterns. To enhance forecast accuracy, a neural network method is employed to capture nonlinear patterns, combined with the Singular Spectrum Analysis (SSA) and Time Series Regression (TSR) methods, forming the hybrid SSA-TSR-NN. The hybrid SSA-TSR-NN method will be compared with the hybrid SSA-TSR-ARIMA based on Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE). SSA is utilized to separate the original time series data into several independent components, including trend, seasonal, and noise components. TSR is used to model trend and noise components, while NN is applied to model seasonal components and the residual TSR noise component. This research is applied to two regions of Bank Indonesia's inflow and outflow data from July 2011 to June 2023. The conclusion drawn from the study is that, in general, SSA-TSR-NN is capable of reducing forecast errors by 27% compared to SSA-TSR-ARIMA. The neurons and inputs used in NN vary for each inflow and outflow data; having more neurons does not always improve model performance. For the forecast of the next year, the highest inflow occurs in January and April, while the highest outflow is observed in December and April.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ARIMA, Hybrid, Inflow, Neural Network, Outflow, Singular Spectrum Analysis, Time Series Regression
Subjects:	H Social Sciences > H Social Sciences (General) > H61.4 Forecasting in the social sciences H Social Sciences > HA Statistics H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HJ Public Finance Q Science > QA Mathematics > QA329.6 Hankel operators. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Anisah Balqis
Date Deposited:	26 Jan 2024 02:23
Last Modified:	26 Jan 2024 02:23
URI:	http://repository.its.ac.id/id/eprint/105655

Actions (login required)

View Item