Prediksi Pergerakan Indikator Makroekonomi terhadap Risiko Kredit berdasarkan Perubahan Iklim

Sierra, Evelyn (2025) Prediksi Pergerakan Indikator Makroekonomi terhadap Risiko Kredit berdasarkan Perubahan Iklim. Masters thesis, Institute Technology Sepuluh Nopember.

Text 6025231091-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Perubahan iklim tidak hanya berdampak langsung pada aktivitas sektor riil, seperti pertanian, manufaktur, konstruksi, perdagangan, dan jasa, tetapi juga mengganggu stabilitas indikator makroekonomi, seperti pertumbuhan PDB, inflasi, dan pengangguran, yang selama ini menjadi variabel utama dalam model prediksi Probability of Default (PD) sebagai bagian dari risiko kredit. Namun, mayoritas pendekatan prediksi gagal bayar saat ini masih bersifat statis, seperti regresi logistik dan decision tree, yang mengasumsikan hubungan tetap antara variabel makroekonomi dan risiko kredit tanpa mampu menangkap dinamika temporal maupun pengaruh faktor eksternal yang berubah seiring waktu. Selain itu, pendekatan tersebut belum mempertimbangkan interdependensi dinamis antara perubahan iklim dan kondisi makroekonomi serta rentan terhadap multikolinearitas dan bias akibat pemilihan variabel yang kurang optimal. Untuk mengatasi keterbatasan tersebut, tesis ini memodelkan hubungan historis antara indikator makroekonomi dan tingkat gagal bayar menggunakan regresi linear optimal berbasis analisis multikolinearitas dan klasterisasi variabel. Selanjutnya, indikator makroekonomi diproyeksikan menggunakan Autoregressive Integrated Moving Average (ARIMA), sedangkan model prediksi PD ditingkatkan dengan Long Short-Term Memory (LSTM) yang mengintegrasikan data banjir sebagai variabel eksogen untuk menangkap dampak perubahan iklim terhadap stabilitas keuangan secara dinamis. Penelitian ini menerapkan metodologi stress testing berbasis LSTM yang dievaluasi selama 100 epoch dan menghasilkan nilai loss akhir sebesar 0,1138. Dengan memanfaatkan dataset makroekonomi dan perubahan iklim yang komprehensif, penelitian ini mengembangkan model prediksi yang mampu menangkap hubungan nonlinier serta interaksi kompleks antarvariabel. Hasil penelitian ini diharapkan dapat berkontribusi pada pengembangan metode penilaian risiko kredit yang lebih tangguh dan akurat dalam menghadapi perubahan iklim sehingga mendukung terjaganya stabilitas sistem keuangan di masa depan.
===============================================================================================================================
Climate change not only directly affects real-sector activities, such as agriculture, manufacturing, construction, trade, and services, but also disrupts the stability of key macroeconomic indicators, including GDP growth, inflation, and unemployment, which have long served as the primary variables in Probability of Default (PD) prediction models for credit risk assessment. However, most existing PD prediction approaches remain static, such as logistic regression and decision trees, assuming fixed relationships between macroeconomic variables and credit risk while failing to capture temporal dynamics and evolving external shocks. Furthermore, these approaches do not account for the dynamic interdependence between climate change and macroeconomic conditions and are susceptible to multicollinearity and bias resulting from suboptimal variable selection. To address these limitations, this thesis models the historical relationship between macroeconomic indicators and default rates using an optimal linear regression approach based on multicollinearity analysis and variable clustering. The macroeconomic indicators are subsequently forecast using the Autoregressive Integrated Moving Average (ARIMA) model, while the PD prediction model is enhanced using a Long Short-Term Memory (LSTM) network that incorporates flood data as an exogenous variable to dynamically capture the impacts of climate change on financial stability. The study applies an LSTM-based stress testing methodology, evaluated over 100 epochs, achieving a final loss value of 0.1138. By utilizing comprehensive macroeconomic and climate-related datasets, this research develops a prediction model capable of capturing nonlinear relationships and complex interactions among variables. The findings are expected to contribute to the development of more robust and accurate credit risk assessment methods in the context of climate change, thereby supporting the long-term stability of the financial system.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Long Short-Term Memory, Makroekonomi, Perubahan Iklim, Risiko Kredit, Climate Change, Credit Risk, Long Short-Term Memory, Macroeconomy
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. H Social Sciences > HB Economic Theory > Economic forecasting--Mathematical models.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Evelyn Sierra
Date Deposited:	04 Aug 2025 06:33
Last Modified:	01 Jul 2026 05:54
URI:	http://repository.its.ac.id/id/eprint/124889

Actions (login required)

View Item