Perbandingan model LSTM dan Transformer Attention Decoder dalam prediksi harga Cryptocurrency

Fauzi, Muhammad Zidan (2026) Perbandingan model LSTM dan Transformer Attention Decoder dalam prediksi harga Cryptocurrency. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

Preview	Text 01. Cover.pdf Download (129kB) | Preview
Preview	Text 02. Abstrak.pdf Download (160kB) | Preview
Preview	Text 03. Keterangan bebas plagiarism.pdf Download (122kB) | Preview
Preview	Text 04. Daftar Isi.pdf Download (238kB) | Preview
Preview	Text 05. Bab 1.pdf Download (223kB) | Preview
	Text 06. Bab 2.pdf Restricted to Registered users only Download (313kB) | Request a copy
	Text 07. Bab 3.pdf Restricted to Registered users only Download (307kB) | Request a copy
	Text 08. Bab 4.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text 09. Bab 5.pdf Restricted to Registered users only Download (140kB) | Request a copy
	Text 10. Daftar Pustaka.pdf Restricted to Registered users only Download (143kB) | Request a copy

Abstract

INDONESIA; Penelitian ini merupakan pengembangan dari studi sebelumnya mengenai prediksi harga aset kripto, dengan fokus pada perbandingan dua model deep learning, yaitu Long Short-Term Memory (LSTM) dan Transformer Attention Decoder. Latar belakang penelitian didasari oleh tingginya volatilitas cryptocurrency, khususnya sektor memecoin, yang jarang menjadi focus penelitian meskipun memiliki potensi pasar yang besar. Data harga memecoin diperoleh dari CoinMarketCap, diproses melalui tahap seleksi dan normalisasi, lalu dimodelkan menggunakan kerangka kerja CRISP-DM. Kedua model dievaluasi dengan Mean Absolute Percentage Error (MAPE) dan Root Mean Square Error (RMSE). Hasil penelitian menunjukkan bahwa kedua model mampu memprediksi harga dengan tingkat akurasi berbeda, dan model dengan error terendah lebih unggul. Temuan ini diharapkan dapat menjadi dasar pengembangan Decision Support System (DSS) yang membantu investor kripto dalam mengambil keputusan berbasis data pada aset dengan volatilitas tinggi. ENGLISH: This research is a continuation of previous studies on cryptocurrency price prediction, focusing on the comparison between two deep learning models, Long Short-Term Memory (LSTM) and Transformer Attention Decoder. The study is motivated by the high volatility of cryptocurrencies, particularly in the memecoin sector, which has rarely been the focus of research despite its significant market potential. Price data of memecoins were obtained from CoinMarketCap, processed through selection and normalization stages, and then modeled using the CRISP-DM framework. Both models were evaluated using Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). The results show that both models are capable of predicting prices with varying levels of accuracy, with the model producing the lower error performing better. These findings are expected to serve as the foundation for developing a Decision Support System (DSS) to assist crypto investors in making data-driven decisions on highly volatile assets.

Actions (login required)

View Item