Prediksi Pergerakan Harga Pada Cryptocurrency Dengan Menggunakan Metode Hybrid SARIMAX-LSTM

Achmadi, Galih Ridha (2024) Prediksi Pergerakan Harga Pada Cryptocurrency Dengan Menggunakan Metode Hybrid SARIMAX-LSTM. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025202010-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (5MB) | Request a copy

Abstract

Nama besar teknologi blockchain muncul beriringan dengan mata uang digital paling fenomenal yaitu Bitcoin (BTC). Dengan adanya blockchain, digital assets semakin pesat berkembang. Menurut Coinmarketcap per Maret 2021, terdapat lebih dari 5.000 jenis cryptocurrency. Ini adalah jumlah yang sangat banyak dan telah meningkat berkali-kali lipat sejak 2009 silam, saat Bitcoin baru mulai diperkenalkan kepada publik. Fenomena ini sangat menarik bagi para investor dan trader, mengingat karakteristik dan fundamental cryptocurrencies sangat mirip dengan stock market (pasar saham). Namun tantangan dalam memprediksi pergerakan harga cryptocurrencies adalah volatilitas mata uang (contoh; BTC-USD) yang sangat tinggi. Oleh karena itu, dalam studi ini dibuat sebuah metode prediksi yang lebih akurat terhadap pergerakan harga cryptocurrency tersebut. Pentingnya prediksi pergerakan harga cryptocurrency adalah untuk memperkirakan harga pada masa yang akan datang dimana karakteristik volatilitas harganya yang sangat tinggi dibandingkan instrumen keuangan yang lain seperti pasar saham dan valas. Metode pada penelitian ini mengembangkan model linier seasonal Hybrid SARIMAX-LSTM (Seasonal Auto-Regressive Integrated Moving Average with eXogenous factors – Long Short Term Memory) untuk membandingkan nilai prediksi dari pemodelan tunggal ARIMA (Auto-Regressive Integrated Moving Average) dan LSTM (Long Short Term Memory). Ketiga metode tersebut digunakan untuk mencari nilai Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) dan Mean Absolute Error (MAE) yang terkecil, sehingga dapat ditentukan metode mana yang memiliki performansi terbaik. Dari penelitian ini diharapkan pemodelan secara Hybrid SARIMAX-LSTM memiliki nilai akurasi terbaik dan error terkecil dibandingkan dengan pemodelan tunggal.
=================================================================================================================================
In today's digital era, blockchain technology has become a major topic of discussion across many industry sectors, especially in the financial world. Blockchain, which underpins cryptocurrency, has witnessed rapid growth since the introduction of the first digital currency, Bitcoin (BTC), in 2009. This unique technology enables transparent, secure, and decentralized transactions. As of September 2023, data from Coinmarketcap indicates that there are over 9,000 types of cryptocurrencies circulating in the global market, a phenomenal growth from a digital currency concept that was once known to only a few. The cryptocurrency market, with all its potential rewards, comes with the risk of high price volatility. This presents a distinct challenge for investors and traders looking to capitalize on this market. Addressing this challenge, there's an urgent need for reliable and accurate prediction methods. This research aims to develop an accurate cryptocurrency price prediction method to assist market participants in making informed investment decisions and minimizing the risk of losses. As a solution, this research combines two proven modeling approaches from various fields, namely ARIMA and LSTM, into a model known as the Hybrid SARIMAX-LSTM. A key aspect of this model is the variable of "volume" as an exogenous factor. Trading volume is often viewed as a vital indicator of market interest and activity, and by incorporating it into the model, this research strives to enhance the accuracy of cryptocurrency price predictions. Using evaluation metrics like MSE, RMSE, MAPE, and MAE, this research compares the performance of the Hybrid SARIMAX-LSTM model with single ARIMA and LSTM models to determine the most accurate method. Preliminary results from this research suggest that the Hybrid SARIMAX-LSTM model, considering the exogenous factor "volume," offers the potential for more accurate predictions with lower error rates compared to single modeling approaches. This indicates that the hybrid approach could be a solution to the challenges of volatility in the cryptocurrency market. With more accurate prediction information, investors and traders can make better decisions, maximizing profits and minimizing risks.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	prediksi, data time series, SARIMAX, cryptocurrency, bitcoin, prediction, time series data, ARIMA, neural network, cryptocurrency
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > Q Science (General) > Q337.5 Pattern recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Galih Ridha Achmadi
Date Deposited:	10 Feb 2024 02:18
Last Modified:	10 Feb 2024 02:18
URI:	http://repository.its.ac.id/id/eprint/106426

Actions (login required)

View Item