Gede, Fajar Kananda A.W. (2026) Model Pengukuran Performa Finansial dan Pengambilan Keputusan Berbasis Algoritma LSTM - CAMELS pada Lembaga Keuangan Non-Bank. Masters thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 6010242015-Master_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
6010242015-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Lembaga Perkreditan Desa (LPD) merupakan lembaga keuangan berbasis adat yang berperan penting dalam mendukung perekonomian masyarakat Bali. Namun, lemahnya pengawasan dan tingginya risiko operasional menyebabkan sebagian LPD rentan mengalami penurunan kesehatan keuangan. Metode penilaian yang digunakan saat ini masih bergantung pada rasio keuangan yang merupakan lagging indicator, sehingga pemantauan bersifat retrospektif dan belum mampu memberikan peringatan dini terhadap potensi penurunan kinerja di masa mendatang. Penelitian ini bertujuan mengembangkan model pengukuran kesehatan keuangan prediktif berbasis integrasi Long Short-Term Memory (LSTM) dan Capital Adequacy, Asset Quality, Management Quality, Earnings, Liquidity, and Sensitivity (CAMELS) serta mengimplementasikannya dalam sistem pengambilan keputusan untuk mendukung pengawasan dan pengambilan keputusan pada LPD. Penelitian ini menggunakan data laporan keuangan bulanan LPD periode 2010–2025 untuk memprediksi tujuh rasio CAMELS, yaitu Capital Adequacy Ratio (CAR), Kualitas Aktiva Produktif (KAP), Cadangan Pinjaman Ragu-Ragu (CPRR), Return on Assets (ROA), Biaya Operasional – Pendapatan Operasional (BOPO), Likuiditas, dan Loan to Deposit Ratio (LDR). Hyperparameter model ditentukan menggunakan Autocorrelation Function (ACF) dan bayesian optimization dengan tools Optuna. Hasil prediksi kemudian diintegrasikan ke dalam sistem yang mampu menghitung indeks kesehatan, mengidentifikasi potensi pelanggaran ambang regulasi, serta menghasilkan alternatif tindakan perbaikan. Hasil penelitian menunjukkan bahwa meskipun aset dan permodalan menunjukan kondisi yang kuat, LPD mengalami penurunan profitabilitas yang salah satunya disebabkan oleh penyaluran kredit yang lebih konservatif pascapandemi Covid-19. Model LSTM yang dikembangkan mampu menghasilkan prediksi yang akurat untuk sebagian besar rasio, dengan performa terbaik pada CAR dan BOPO serta kemampuan yang baik dalam menangkap pola musiman pada ROA. Sistem pengambilan keputusan yang dihasilkan mampu untuk melakukan peramalan rasio dan indeks kesehatan, komparasi dengan standar minimum, serta menghasilkan opsi perbaikan. Model dan sistem pengambilan keputusan ini memiliki potensi dalam membantu regulator dan pengelola LPD untuk mengidentifikasi risiko keuangan sejak dini, memprediksi kondisi kesehatan pada periode mendatang, serta merumuskan alternatif tindakan perbaikan yang lebih tepat berdasarkan proyeksi rasio keuangan dan dampaknya terhadap kesehatan LPD secara keseluruhan.
=======================================================================================================================================
Lembaga Perkreditan Desa (LPD) are traditional community-based financial institutions that play an important role in supporting the economy of Bali. However, weak oversight and high operational risks make some LPDs vulnerable to financial deterioration. Current assessment methods still rely on financial ratios that serve as lagging indicators, resulting in retrospective monitoring that cannot adequately provide early warnings of potential future performance deterioration. This study aims to develop a predictive financial health assessment model by integrating Long Short-Term Memory (LSTM) and Capital Adequacy, Asset Quality, Management Quality, Earnings, Liquidity, and Sensitivity (CAMELS), and to implement the model in a decision support system to facilitate LPD supervision and decision-making. This study uses monthly LPD financial statements from 2010 to 2025 to predict seven CAMELS ratios: Capital Adequacy Ratio (CAR), Credit Quality (KAP), Non-Performing Loan Reserves (CPRR), Return on Assets (ROA), Operating Expenses to Operating Income (BOPO), Liquidity Ratio, and Loan to Deposit Ratio (LDR). Model hyperparameters are determined using the Autocorrelation Function (ACF) and Bayesian optimization with the Optuna library. The resulting predictions are integrated into a system capable of calculating financial health indices, identifying potential breaches of regulatory thresholds, and generating alternative corrective actions. The results indicate that although LPD assets and capital positions remain relatively strong, profitability has declined, partly due to more conservative lending practices following the COVID-19 pandemic. The developed LSTM model produces accurate predictions for most ratios, with the best performance achieved for CAR and BOPO, while also effectively capturing seasonal patterns in ROA. The resulting decision support system can forecast financial ratios and health indices, compare them against minimum standards, and generate corrective action alternatives. This model and decision support system have the potential to help regulators and LPD managers identify financial risks at an early stage, predict future financial health conditions, and formulate more appropriate corrective actions based on projected financial ratios and their implications for overall LPD financial health.
Actions (login required)
 |
View Item |